printmgr file

• The Hong Kong Public Offering has transitioned to a fully electronic application process, eliminating the distribution of printed prospectus copies.
• Investors must access the prospectus digitally via the HKEXnews website or the company's official website at www.zhipuai.cn.
• Applications can only be submitted through two digital channels: the HK eIPO White Form service or the HKSCC EIPO channel via brokers.
• A strict minimum application of 100 Hong Kong Offer Shares is required, with further increments following specific tiered multiples.
• The maximum individual application is capped at 935,500 shares, representing 50% of the initial Hong Kong Offer Shares.
• Full payment or pre-funding is mandatory at the time of application, with costs ranging from approximately HK$11,737 to over HK$109 million.

We will not provide any physical channels to accept any application for the Hong Kong Offer Shares by the public.

• The total payment for shares includes mandatory fees such as brokerage, SFC transaction levies, and Stock Exchange trading fees.
• The Hong Kong Public Offering is scheduled to commence on December 30, 2025, and conclude its application phase by January 5, 2026.
• Strict adherence to the specified number of shares is required, as non-compliant applications are subject to immediate rejection.
• Allocation results and interest levels will be disclosed across multiple digital platforms and a dedicated telephone enquiry line by January 7, 2026.
• Successful applicants will receive H Share certificates or have them deposited into the CCASS system by the final settlement date.

No application for any other number of Hong Kong Offer Shares will be considered and any such application is liable to be rejected.

• The document outlines the critical dates for the public offering, including the commencement of H Share trading on January 8, 2026.
• Strict deadlines are established for the HK eIPO White Form service, with application submissions closing at 11:30 a.m. and payments at 12:00 noon.
• Contingency plans are in place for severe weather, such as 'black' rainstorm warnings or tropical cyclone signals, which may delay the opening of application lists.
• H Share certificates only become valid evidence of title at 8:00 a.m. on the Listing Date, provided the Global Offering becomes unconditional.
• Refunds for unsuccessful or partially successful applications are scheduled for dispatch on or before the trading commencement date.
• Investors are warned that trading based on allocation details prior to the validation of share certificates is done entirely at their own risk.

Investors who trade the H Shares on the basis of publicly available allocation details prior to the receipt of H Share certificates or prior to the H Share certificates becoming valid evidence of title do so entirely at their own risk.

• Applicants for 500,000 or more shares via HK eIPO White Form must collect certificates in person at the Far East Finance Centre.
• Individual applicants are strictly prohibited from authorizing third parties to collect certificates on their behalf.
• Refunds are processed via e-Auto Refund for single bank account users, while multiple bank account users receive physical checks by post.
• Small-scale applicants (under 500,000 shares) will receive all documentation via ordinary post at their own risk.
• The Global Offering is contingent upon becoming unconditional; if terminated, the company will issue a formal announcement.
• This prospectus is legally restricted to the Hong Kong Public Offering and does not constitute an offer in any other jurisdiction.

Applicants being individuals who are eligible for personal collection may not authorize any other person to collect on their behalf.

• Investors are strictly cautioned to rely only on the information explicitly provided within the prospectus document.
• The company disclaims any responsibility for unauthorized information or representations made by third parties, including underwriters and directors.
• Official digital content on the company's website is explicitly excluded from forming part of the legal prospectus.
• The document outlines a comprehensive structure covering risk factors, regulatory overviews, and detailed financial information.
• A significant portion of the document is dedicated to the business operations, history, and corporate structure of Zhipu AI.

Information contained on our website, located at www.zhipuai.cn, does not form part of this prospectus.

• The document outlines the structure and legal appendices for a specialist technology company's initial public offering on the Hong Kong Stock Exchange.
• The company is seeking a listing under Chapter 18C of the Listing Rules, a specific pathway for technology companies that do not meet standard financial requirements.
• Founded in 2019, the firm is a leading Chinese AI developer focused on Artificial General Intelligence (AGI) and the Model-as-a-Service (MaaS) platform.
• The company launched GLM, China’s first proprietary pre-trained large model framework, and open-sourced a 100 billion-scale model in 2022.
• Market data shows significant growth with a revenue CAGR of over 130% between 2022 and 2024, capturing a 6.6% market share in China's general-purpose large model sector.

In particular, we are a specialist technology company seeking to list on the Main Board of the Hong Kong Stock Exchange under Chapter 18C of the Listing Rules because we are unable to meet the requirements under Rule 8.05(1), (2) or (3) of the Listing Rules.

• The company ranks as China's largest independent general-purpose large model developer by revenue as of 2025.
• Their business model centers on a Model-as-a-Service (MaaS) platform offering language, multimodal, agent, and coding models.
• The platform demonstrates significant growth with a revenue CAGR exceeding 130% from 2022 to 2024.
• Global reach is evidenced by over 45 million downloads of open-source models and empowerment of approximately 80 million devices.
• The MaaS architecture is built on three pillars: a comprehensive model portfolio, scalable applications, and easy infrastructure adaptability.
• The GLM-4.6 model achieved top-tier global rankings for coding capabilities on the CodeArena evaluation platform in late 2025.

Through this product development and commercialization platform, we deliver intelligence to institutional customers, developers and individual customers in the most suitable, sensible and scalable way despite great heterogeneity in computing infrastructure, devices and applications.

• The Model-as-a-Service (MaaS) platform supports a wide range of model sizes from 1.5 billion to 230 billion parameters for diverse hardware.
• On-premise deployment provides organizations with greater control over data security and performance optimization for specialized domains.
• Cloud-based deployment offers agility and cost-effectiveness by eliminating the need for local infrastructure and utilizing usage-based pricing.
• The AI development strategy focuses on three core human faculties: deep thinking, cognition, and tool use.
• The GLM series foundation models serve as the base for specialized reflection, multimodal, agent, and coding models.
• Revenue recognition varies by deployment type, occurring at delivery for on-premise solutions and over the contract term for cloud services.

To teach machines to think like humans, we must empower AI with three core human faculties: deep thinking, cognition and tool use.

• GLM-4.5 is a 355-billion-parameter model that excels in agentic, reasoning, and coding tasks while supporting multi-modal extensions.
• The model ranked first among global open-source models and first in China across twelve industry-standard benchmark tests in July 2025.
• GLM-4.5 achieved the world's second-lowest hallucination rate in RAG evaluations, indicating high reliability in avoiding non-existent answers.
• Market adoption is significant, with token consumption on OpenRouter consistently ranking among the top ten globally and top three for Chinese companies.
• The product line includes specialized versions like GLM-4.5-Air for efficiency and the GLM-Z1 series for 'deep thinking' and complex reasoning.

Reflection and rumination models spend additional time “deep thinking” before generating an answer, which makes them better for complex reasoning tasks.

• GLM-Z1-Rumination introduces iterative, deep-thinking capabilities designed to solve open-ended problems through external information gathering.
• The multimodal suite spans image, video, and voice generation, including GLM-Realtime for live video interactions.
• AutoGLM marks a shift from conversational AI to 'acting' AI, capable of autonomously controlling digital devices via graphical user interfaces.
• AutoGLM 2.0 operates on the cloud, allowing the AI to complete tasks across mobile apps and websites without interrupting the user's local device usage.
• The company leverages its position as the first in China to develop 100-billion-parameter models to build a comprehensive Model-as-a-Service (MaaS) platform.
• Strategic focus remains on scaling general-purpose R&D and fostering a vibrant open-source ecosystem to drive commercialization.

AutoGLM represents a major step forward in the evolution of our AI universe—from “chat” to “act,” bridging the gap between conversation-based AI and real-world task execution.

• The company is officially classified as a Specialist Technology Industry provider under the Artificial Intelligence subcategory of Next-generation Information Technology.
• Their core offerings include multimodal large models capable of processing text, images, audio, and video through billions of parameters.
• Commercial applications span diverse sectors including municipal transportation management, financial services, and large-scale document analysis for internet firms.
• The China Large Language Model (LLM) market is projected to experience explosive growth, rising from RMB5.3 billion in 2024 to RMB101.1 billion by 2030.
• Institutional and enterprise customers are identified as the primary drivers of this market, expected to maintain a 63.7% compound annual growth rate.
• The regulatory landscape currently requires no specific approvals for the company's specialist technology products as of the latest practicable date.

A large model is a complex set of algorithms that identify and leverage patterns from vast amounts of data by iteratively adjusting billions of parameters to increase accuracy.

• The Chinese LLM market is divided between 'non-independent' technology giants and 'independent' pure-play providers.
• Tech giants leverage massive existing user bases and diversified business lines to promote their AI products rapidly.
• Enterprise clients often prefer independent providers to avoid direct competition with tech giants or falling into their spheres of influence.
• Market data from 2024 shows a mix of independent and non-independent firms among the top five revenue earners, with the leader holding a 14.9% market share.
• The featured company emphasizes a research-driven culture, employing 657 R&D specialists focused on natural language processing and multimodal analysis.
• Intellectual property protection is a core strategic pillar, evidenced by 86 registered patents and hundreds of pending applications.

Also, enterprise clients in certain industries can be sensitive about falling or appearing to fall into spheres of influence by certain technology giants and are more inclined to adopt AI solutions from “pure-play” providers.

• The company secures intellectual property through strict ownership clauses with employees and consultants, reporting no material infringement disputes to date.
• The Large Language Model (LLM) market is characterized by intense competition based on technical capabilities, ecosystem building, and talent acquisition.
• Revenue has demonstrated explosive growth with a CAGR exceeding 130%, rising from RMB57.4 million in 2022 to RMB312.4 million in 2024.
• Significant net losses have occurred alongside revenue growth, primarily driven by massive investments in research and development which reached RMB2,195.4 million in 2024.
• The path to profitability relies on leveraging the expanding Chinese LLM market, iterating AI models, and broadening customer reach to achieve operating leverage.
• Future sustainability is predicated on the belief that revenue growth will eventually outpace the high costs of R&D and infrastructure.

Our losses during the Track Record Period were primarily due to our significant investments in research and development.

• The company is prioritizing profitability through optimized R&D efficiency and the iteration of its foundation models via its MaaS platform.
• Operational management strategies include simplifying workflows, controlling administrative expenses, and adopting data-driven marketing approaches.
• Customer concentration is high but declining, with the top five customers accounting for 40.0% of revenue in mid-2025 compared to 61.5% in 2023.
• The supply chain relies heavily on providers of computing hardware, data cleansing services, and large model evaluation support.
• A group of controlling shareholders, including several doctors and investment entities, maintains a collective interest of approximately 33.03%.
• The company asserts its business model is sustainable based on historical growth and strategic cloud-based deployment.

With our deep academic roots as cornerstone for technological leadership, we are constantly making technological advancements and iterations of our foundation models, boosting our profitability.

• The company is applying for a Hong Kong Stock Exchange listing under Rule 18C.03, targeting a market capitalization exceeding HK$6 billion.
• Pre-IPO funding rounds have successfully raised over RMB8,360 million across eight distinct investment phases.
• The business faces significant R&D risks, specifically the necessity of continuous innovation in the rapidly changing AI industry to remain competitive.
• Operational dependencies include a heavy reliance on third-party computing resources and the performance of proprietary Model-as-a-Service (MaaS) platforms.
• Regulatory and market uncertainties persist, particularly regarding evolving Chinese AI laws and the early-stage development of Artificial General Intelligence (AGI).
• The company admits to a limited track record in commercialization, making future demand and market size difficult to predict accurately.

The development of AGI is still at an early stage and there are substantial uncertainties in the future realization of AGI.

• The company experienced rapid revenue growth, increasing from RMB 57.4 million in 2022 to RMB 312.4 million in 2024.
• Gross profit margins remained relatively stable between 48.9% and 64.6% throughout the reported periods.
• Research and development expenses are the primary driver of losses, reaching over 700% of total revenue in 2024.
• Operating losses have widened significantly, totaling over RMB 2.5 billion in 2024 as the company scales its operations.
• Financial statements include significant non-cash adjustments, such as changes in the carrying amounts of financial instruments issued to investors.
• The company remains in a pre-profit stage with total losses for the first half of 2025 exceeding RMB 2.3 billion.

Research and development expenses . . . . . . . . . . (84,377) (147.0) (528,884) (424.7) (2,195,436) (702.7)

• The company utilizes 'adjusted loss for the year' as a non-IFRS measure to facilitate period-to-period comparisons of operating performance.
• Adjustments to the net loss include adding back equity-settled share-based compensation, changes in the carrying amount of financial instruments, and listing expenses.
• Management argues these adjustments provide useful information by removing non-cash items and one-time costs like those associated with the Global Offering.
• Financial data shows a significant increase in net loss from 2022 to 2025, with the adjusted net loss for the first half of 2025 reaching over 1.7 billion RMB.
• Revenue is primarily driven by on-premise deployment of large model services through their MaaS platform, consistently outperforming cloud-based deployment.

The use of such non-IFRS measure has limitations as an analytical tool, and you should not consider them in isolation from, or as substitute for analysis of, our consolidated financial statements.

• The company began generating revenue from overseas on-premise deployments in 2024, with a significant initial focus on Southeast Asian markets like Malaysia and Singapore.
• While Mainland China remains the dominant revenue source, the share of on-premise revenue from Southeast Asia grew to 11.1% by the first half of 2025.
• Cost of revenue is primarily driven by payroll for service personnel and computing service fees, the latter of which saw a sharp increase in 2024 and 2025.
• The company utilizes third-party outsourcing for labor-intensive tasks such as data annotation to manage technical service and consultation costs.
• Gross profit margins for on-premise deployment have remained relatively strong, though cloud-based deployment margins show significant volatility and a downward trend.

Starting from 2024, our large model on-premise deployment services have generated revenue from overseas customers, primarily from customers in Southeast Asia.

• The company experienced a dramatic increase in net losses, rising from RMB143.7 million in 2022 to over RMB2.9 billion by 2024.
• Significant capital is being deployed into research and development to build advanced foundation models and expand the R&D team.
• Operating costs are further driven by the procurement of massive computing services and hardware from third parties to support business expansion.
• Aggressive sales and marketing investments are being made to capture emerging market opportunities and grow the customer base.
• The financial position shows a widening net liability gap, reaching RMB6.15 billion by mid-2025, largely due to comprehensive losses and financial instruments issued to investors.

Such overall upward trend in our net loss was primarily due to (i) the significant increase in research and development expenses as we expanded our R&D team and procured substantially more computing services from third parties and related computing hardware.

• Net current liabilities increased significantly to RMB7,088.9 million by mid-2025, largely due to financial instruments issued to Pre-IPO investors.
• The company expects to transition from a net liability position to a net asset position following the Global Offering as investor rights terminate and liabilities are re-designated as equity.
• Operating activities show a consistent and widening net cash outflow, reaching over RMB1.3 billion for the first half of 2025 alone.
• The monthly cash burn rate has escalated dramatically from RMB3.0 million in 2022 to RMB327.3 million in 2025, driven by R&D and hardware procurement.
• Despite high burn rates, the company maintains substantial liquidity with RMB8,943.1 million in cash and facilities as of October 2025, supplemented by expected IPO proceeds.

Our historical cash burn rate was RMB3.0 million, RMB105.9 million, RMB194.5 million and RMB327.3 million for each of the years ended December 31, 2022, 2023 and 2024 and for the six months ended June 30, 2025, respectively.

• The company estimates a financial runway of 27.3 to 38.9 months depending on the utilization of net proceeds from its upcoming listing.
• Operational sustainability is predicated on maintaining current employee levels in R&D and avoiding significant capital investments or fixed asset acquisitions.
• Net cash used in operating activities has increased dramatically year-over-year, rising from RMB68.2 million in 2022 to over RMB1.3 billion in the first half of 2025 alone.
• The company's liquidity position shows signs of strain, with the current ratio declining steadily from 0.6 in 2022 to 0.3 by mid-2025.
• Significant losses before tax are the primary drivers of cash outflows, though these are partially offset by changes in financial instruments and trade payables.
• The global offering is priced at HK$116.20 per share, targeting a total market capitalization of approximately HK$51.155 billion.

We will continue to monitor our cash flows from operations closely and maintain our financial viability through a variety of means, including, among others, banking facilities and external financings.

• The company expects to receive net proceeds of approximately HK$4,173.4 million based on an offer price of HK$116.20 per share.
• A significant majority of the proceeds, approximately 70%, is earmarked for strengthening research and development in general-purpose large AI models.
• The remaining funds are split equally between optimizing the Model-as-a-Service (MaaS) platform, expanding business partner networks, and general working capital.
• The unaudited pro forma adjusted net tangible assets per share is calculated at HK$17.85 based on over 440 million shares in issue.
• The company does not intend to pay cash dividends in the foreseeable future, opting instead to retain earnings for business expansion.
• Future dividend payments are subject to director discretion and strict PRC regulatory requirements regarding after-tax profits.

Investors should not purchase our ordinary shares with the expectation of receiving cash dividends.

• The company anticipates it will be unable to pay dividends in the foreseeable future due to significant accumulated losses and regulatory reserve requirements.
• Total listing expenses are estimated at RMB158.5 million, representing approximately 4.0% of the gross proceeds from the Global Offering.
• The company and nine subsidiaries were added to the U.S. Entity List in early 2025, restricting access to certain American software and technology.
• Management believes the Entity List designation will not have a material adverse impact as they do not currently rely on items subject to Export Administration Regulations.
• Recent technological advancements include the release of flagship foundation models GLM-4.5 and GLM-4.5V, focusing on reasoning and visual comprehension.

In light of our accumulated losses as disclosed in this prospectus, it is unlikely that we will be eligible to pay dividends out of our profits in the foreseeable future.

• The company introduced AutoGLM, an autonomous agent powered by GLM-4.5 that simulates human actions on mobile and web applications without occupying the user's device.
• Institutional customer growth surged to over 12,000 by September 2025, with daily token consumption reaching 4.2 trillion in November 2025.
• Revenue for the third quarter of 2025 is estimated to increase by over 60% year-over-year, though net losses are expected to rise due to heavy R&D spending.
• The company confirmed that the COVID-19 pandemic had no material adverse impact on its operations, noting a 116.9% revenue increase between 2022 and 2023.
• Regulatory milestones were achieved with the completion of the CSRC filing in December 2025 for the company's global offering.

It can autonomously complete requested tasks on the cloud without occupying the user’s mobile phone or computer, allowing users to continue using their own devices without interruption.

• This section provides a formal glossary of terms and legal entities essential for understanding the company's prospectus.
• It details a series of 'Beijing Knowledge' subsidiaries, most of which were established between 2023 and 2025 as wholly owned entities.
• The text defines the geographical scope of 'China' or 'PRC' to specifically exclude Hong Kong, Macau, and Taiwan for the purposes of the document.
• Operational definitions are provided for financial infrastructure, including the Central Clearing and Settlement System (CCASS) and the role of Capital Market Intermediaries.
• The document identifies Beijing Lianpai as a limited partnership and one of the company's Controlling Shareholders.

“Beijing Knowledge Qingying” Beijing Knowledge Qingying Culture Media Co., Ltd. (北京智譜清影科技文化傳媒有限公司), a limited liability company established under the laws of the PRC on March 13, 2025, and a wholly owned subsidiary of our Company;

• The document defines Knowledge Atlas Technology Joint Stock Company Limited, a PRC-based entity established in 2019 and converted to a joint stock company in 2025.
• It identifies a core group of controlling shareholders including Dr. Liu, Dr. Tang, Dr. Li, Dr. Xu, and Dr. Zhang, who are bound by a Concert Party Agreement.
• The text outlines the conversion of over 178 million unlisted shares into H Shares for listing on the Stock Exchange following CSRC filing.
• Key regulatory and operational terms are defined, including the PRC Company Law, the Entity List administered by BIS, and the FINI online platform for new listings.
• Provisions are included for 'Extreme Conditions,' such as super typhoons, that might affect the public's ability to work or impact safety in Hong Kong.

“Extreme Conditions” extreme conditions as announced by the government of Hong Kong in the case where a super typhoon or other natural disaster of a substantial scale seriously affects the working public’s ability to resume work or brings safety concern for a prolonged period;

• This section provides formal definitions for key entities and financial terms involved in a Global Offering on the Hong Kong Stock Exchange.
• It details the structure of the Hong Kong Public Offering, including specific share counts and associated transaction fees like the SFC levy and Stock Exchange trading fee.
• The document identifies specific corporate subsidiaries such as Hangzhou Knowledge Atlas and Huangshi Knowledge Atlas, establishing their legal status within the Group.
• Technical procedures for share applications are defined, specifically outlining the HK eIPO White Form and HKSCC EIPO systems for digital and nominee-based subscriptions.
• The text establishes the legal framework for the offering by referencing the Guide for New Listing Applicants and the Hong Kong Underwriting Agreement.

“Hong Kong Public Offering” the offer for subscription of the Hong Kong Offer Shares to the public in Hong Kong at the Offer Price (plus brokerage of 1.0%, SFC transaction levy of 0.0027%, Stock Exchange trading fee of 0.00565% and AFRC transaction levy of 0.00015%).

• The document defines Zhuhai Hengqin Huihui as a PRC-established limited partnership serving as an Employee Ownership Platform and Controlling Shareholder.
• It outlines the International Offering parameters, involving over 35 million H Shares targeted at non-U.S. investors under Regulation S.
• The text establishes the legal framework for International Sanctions, referencing regulations from the U.S., EU, UK, UN, and Australia.
• Key corporate subsidiaries are identified, including Jincheng Yaoda in the British Virgin Islands and Jingsheng Hengxing in Singapore.
• The timeline for the public offering is specified, with a Latest Practicable Date in late 2025 and an expected Listing Date in early January 2026.

“International Sanctions” all applicable laws and regulation to economic sanctions, export controls, trade embargoes and wider prohibitions and restrictions on international trade and investment related activities.

• The document defines key financial terms and regulatory bodies governing a public offering on the Stock Exchange of Hong Kong.
• It identifies 'Mr. Wang' (Wang Shaolan) as a co-founder and deputy general manager of the company.
• The 'Offer Price' is set at HK$116.20 per share, excluding various transaction levies and brokerage fees.
• The company maintains several wholly owned subsidiaries under the 'Knowledge Atlas' and 'Knowledge Huanyu' branding across Nanjing, Shanghai, and Shenzhen.
• A 'Share Subdivision' is detailed, converting each single share of RMB1.00 into ten shares of RMB0.10 each prior to listing.
• The 'Over-allotment Option' allows for the issuance of up to 5,612,900 additional H Shares to cover potential over-allocations.

“Share Subdivision” the sub-division of the Shares by the Company where the Company subdivided its Shares from one Share of RMB1.00 each into ten Shares of RMB0.10 each, which will become effective immediately prior to the Listing;

• The document defines key legal and financial entities involved in a corporate listing, including the Sole Sponsor and Stabilizing Manager, China International Capital Corporation.
• It outlines specific regulatory categories under Chapter 18C of the Listing Rules, focusing on Specialist Technology companies and products.
• The text details the corporate structure of the Group, identifying various subsidiaries such as Tianjin Knowledge Atlas and Zhejiang Knowledge Xinpian.
• A significant portion of the definitions addresses U.S. regulatory compliance, specifically citing Executive Order 14105 and the final rule effective January 2025.
• The Track Record Period for the financial reporting is defined as the three years ending December 31, 2024, plus the first half of 2025.
• The definitions clarify the status of Unlisted Shares and the roles of Employee Ownership Platforms like Zhideng in the company's control structure.

“U.S. Investors” (i) U.S. persons or (ii) United States person as defined in 31 CFR Part 850.229, the final rule that implements Executive Order 14105 and became effective in January 2025.

• The text identifies Zhuhai Knowledge Linghang Technology Co., Ltd. as a wholly owned subsidiary established in late 2024.
• A comprehensive list of acronyms outlines the regulatory landscape, including Hong Kong's AFRC and the U.S. Bureau of Industry and Security (BIS).
• Key Chinese governmental bodies such as the CAC, CSRC, and MIIT are defined to establish the legal and administrative context of the company's operations.
• Technical definitions for AI-related terms are provided, distinguishing between standard AI, autonomous AI agents, and the pursuit of Artificial General Intelligence (AGI).
• The document clarifies that in cases of linguistic inconsistency regarding Chinese entities or laws, the Chinese versions shall prevail over English translations.

“AGI” artificial general intelligence, a sophisticated level of artificial intelligence that matches and even surpasses human capabilities across all cognitive tasks;

• The text provides foundational definitions for core AI concepts including algorithms, machine learning, and deep learning architectures.
• It distinguishes between specialized hardware like GPUs and software frameworks like APIs and Integrated Development Environments (IDEs).
• Key linguistic concepts are defined, such as corpora for training data and the distinction between Natural Language Processing (NLP) and Understanding (NLU).
• The glossary addresses the limitations of AI, specifically defining 'hallucination' as the generation of plausible but fabricated information.
• It outlines modern deployment models like Model-as-a-Service (MaaS) and the scaling efficiency of Mixture of Experts (MoE) architectures.

“hallucination” the phenomenon where a large language model generates plausible-sounding but factually incorrect or fabricated information;

• The text defines essential AI terminology including PPO, RLHF, and SOTA performance to establish a technical baseline.
• It explains the structural components of LLMs, such as Transformers, tokens, and the concept of sparse activation for efficiency.
• The document outlines the training pipeline from pre-training to supervised fine-tuning (SFT) and human feedback optimization.
• A formal legal disclaimer identifies forward-looking statements based on management's current beliefs and future expectations.
• The text warns that these projections are subject to significant risks, including market competition, regulatory changes, and financial volatility.
• Specific operational uncertainties are listed, ranging from capital expenditure plans to the ability to retain key management personnel.

These statements are subject to certain risks, uncertainties and assumptions, including the other risk factors as described in this prospectus.

• The company disclaims any obligation to update forward-looking statements regardless of new information or future events.
• Investors are warned that the market price of H Shares could decline significantly, potentially leading to a total loss of investment.
• Material risks are categorized into seven key areas including R&D, commercialization, intellectual property, and jurisdictional challenges.
• The AI industry and Large Language Model (LLM) markets are defined by rapid, constant technological breakthroughs and evolving application modes.
• Unknown or currently immaterial risks may still pose significant threats to the company's future financial condition and operations.

In any such case, the market price of our H Shares could decline, and you may lose all or part of your investment.

• The company's future success depends on its ability to integrate multimodal advancements and AI agents into its large language models.
• Established competitors with existing product portfolios may possess more resources and commercialization experience than independent providers.
• Significant R&D investment is required to overcome technological challenges and differentiate services through innovative, efficient features.
• Rapid technological shifts or the emergence of alternative technologies could render current models obsolete or unattractive to the market.
• The AI industry remains in a developmental stage characterized by significant uncertainties and massive capital requirements for foundation models.
• Financial performance is heavily tied to the ability to scale R&D spending, which has seen exponential growth from RMB84.4 million to over RMB2 billion.

In addition, the emergence of new or alternative technologies could replace or reduce the demand for our models and solutions, render our technologies obsolete or unattractive.

• The company's R&D expenses have dramatically exceeded total revenue, reaching over 1,900% of revenue in some periods.
• Technological advancement is categorized into five stages of Large Language Models, with the company currently attempting to move beyond the first three.
• There is no guarantee that the final stages of AI development—self-perception and consciousness—are technically achievable.
• Rapid innovation in the AI sector creates a high risk of existing solutions becoming obsolete before development costs are recovered.
• The realization of Artificial General Intelligence (AGI) remains speculative, and failure to achieve it could invalidate the company's massive capital investments.
• Success is heavily dependent on a specialized team of data scientists and engineers in a highly competitive talent market.

However, we cannot guarantee that we will be able to the self-perception stage and consciousness stage.

• The company faces intense competition for AI experts with deep technical backgrounds in natural language processing and complex systems.
• Loss of core scientists would be difficult to replace and could severely disrupt business operations due to their unique industry relationships and reputations.
• Employee retention is heavily dependent on share incentive schemes and maintaining a competitive, inclusive workplace culture.
• The demand for computational resources is growing exponentially as AI models become more complex, leading to a massive surge in service fees.
• Computing service fees have risen to represent over 70% of total research and development expenses as of 2025.
• Reliance on third-party computing providers creates significant risks regarding price fluctuations, hardware shortages, and potential service disruptions.

If one or more of our core scientists were unable or unwilling to continue to contribute their services to us, we may not be able to replace them in a timely manner, or at all.

• External disruptions such as power loss, geopolitical conflicts, and trade sanctions can severely impact computing power supply chains and service stability.
• The commercial success of the business is fundamentally tied to the research, development, and performance of complex AI models.
• Flawed algorithms or biased, inaccurate data can lead to sub-optimal or inappropriate outputs, triggering regulatory scrutiny and reputational damage.
• Improper configuration or failure to provide adequate maintenance services can result in contract terminations and legal claims.
• Scaling operations requires the recruitment and retention of highly specialized personnel, which is not guaranteed as the customer base grows.

If the data used are inaccurate, incomplete, unrepresentative or biased, or if our algorithms are flawed, incorrectly designed or calibrated, our models may generate inaccurate, sub-optimal, unfair or otherwise inappropriate outputs.

• The LLM market is characterized by intense and increasing competition that is expected to lead to rapid industry consolidation.
• Major competitors often possess superior financial, R&D, and marketing resources, along with stronger brand recognition and partner networks.
• Resource-heavy development may favor a small group of leading players who can better secure computing resources and withstand regulatory shifts.
• Pricing pressures resulting from market competition could lead to significant reductions in profitability and market share.
• Customers may choose competitor models even if the company's technology is technically superior or more cost-efficient.
• The actual size and growth rate of the addressable market for LLM solutions remain uncertain and difficult to predict.

In light of these factors, even if our models are more advanced, effective and cost-efficient than those of our competitors, current or potential customers may accept competitors’ models in lieu of ours.

• The evolving nature of the LLM market makes it difficult to accurately predict customer demand or future growth trajectories.
• Company success is not guaranteed by overall market growth if the firm fails to effectively capture specific opportunities.
• The Model-as-a-Service (MaaS) platform's viability depends on technical reliability, pricing competitiveness, and constant innovation.
• Expanding into diverse industry verticals requires specialized solutions and support that the company may struggle to provide.
• Failure to retain existing customers or attract new ones at competitive prices poses a material threat to revenue growth.

Even if the LLM market grows substantially, there is no guarantee that demand for our solutions will correlate with that growth if we fail to effectively pursue such opportunities.

• Failure to meet customer expectations regarding AI solutions can lead to reputational damage and a significant loss of expected revenue.
• The company faces a limited track record in commercializing large models within the nascent and evolving Chinese consumer market.
• Success depends heavily on attracting and retaining a specialized sales force that possesses both deep AI knowledge and extensive industry resources.
• Expanding into new industry verticals introduces risks related to established competitors, lack of industry-specific familiarity, and regulatory compliance.
• New business models and vertical expansions require substantial upfront R&D costs and management resources before generating any revenue.
• Maintaining compatibility with diverse third-party hardware and platforms is essential for remaining competitive in the AI ecosystem.

Our ability to successfully commercialize our business may involve inherent risks, longer timelines and higher costs as the consumer market for large models in China is still at its early stage of development.

• The company's success depends on maintaining model compatibility across diverse hardware and software platforms, currently supporting over 40 global chip platforms.
• Failure to adapt models to evolving hardware and software technologies could significantly harm the company's competitive edge and financial results.
• The Chinese AI industry is facing an increasingly stringent regulatory environment, including specific measures for generative AI services.
• Generative AI providers in China must assume legal responsibilities for content production, network security, and personal information protection.
• Compliance requirements include mandatory security assessments and algorithm filings for services with public opinion or social mobilization capabilities.
• Uncertainty regarding the interpretation and implementation of evolving PRC laws poses significant legal and operational risks to the business.

The interpretation and implementation of existing measures are evolving and the PRC regulatory agencies, including the CAC, may further adopt new laws, regulations, rules, or detailed implementation and interpretation related to the above-mentioned measures, which may negatively affect us.

• The PRC government has rapidly expanded its regulatory framework for cybersecurity and data protection through several major legislative acts since 2017.
• The Cybersecurity Law and Data Security Law impose broad obligations on 'network operators' and entities engaged in data-related activities.
• The Personal Information Protection Law (PIPL) establishes strict rules for the entire lifecycle of personal data processing.
• Compliance is complicated by the evolving and often uncertain interpretation of these laws, creating significant operational risks for technology firms.
• Companies face liability not only for their own actions but also for the potential failures of third-party data vendors to comply with privacy standards.
• Failure to safeguard data or meet regulatory standards can result in severe legal, financial, and reputational damage.

There is no guarantee as to the effectiveness of the measures we have taken to urge or supervise such third-party vendors or public data controllers to abide by applicable cybersecurity, data protection and personal information protection laws and regulations.

• Current data processing technologies have inherent limitations that may prevent full compliance with evolving cybersecurity and personal information protection laws.
• System failures, security breaches, or cyber attacks could damage the company's reputation and lead to significant legal and financial consequences.
• The 2022 CAC Measures mandate cybersecurity reviews for certain entities, particularly those whose data activities might influence national security.
• There is a lack of clear criteria or interpretation regarding what specific data processing activities are deemed to 'influence national security' by regulators.
• The Regulations on Protection of Critical Information Infrastructure define 'CII' as systems in vital sectors like finance and energy that could endanger public interest if compromised.
• While not yet identified as a Critical Information Infrastructure Operator (CIIO), the company remains subject to potential future designation by industry authorities.

However, the CAC Measures provide no further explanation or interpretation for the criteria on determining the risks that “have or could have an influence on national security.”

• The company faces potential classification as a Critical Information Infrastructure Operator (CIIO), which would trigger mandatory national security reviews.
• New Chinese regulations mandate security assessments and certifications for cross-border transfers of personal information and important data.
• Compliance with evolving data protection laws may result in substantial expenses, operational delays, or the need to fundamentally alter business practices.
• Future legislation and re-interpretations of existing laws create an unpredictable regulatory environment that could lead to fines, lawsuits, or public censure.
• Even unfounded concerns regarding data protection could materially damage the company's reputation and financial prospects.

Any such review process could result in increased compliance costs, delays to our business operations or product deployment, and may have a material adverse effect on our business, financial condition and results of operations.

• Ongoing geopolitical tensions, particularly between China and the U.S., create significant uncertainties for international business operations.
• The implementation of substantial new tariffs and retaliatory protectionist measures by various governments threatens to reduce global economic activity.
• Trade disputes contribute to broader macroeconomic instability, including inflationary pressures and foreign exchange volatility.
• The dynamic and unpredictable trade landscape may lead to increased sanctions, export controls, and investment restrictions targeting high-tech sectors.
• Prolonged economic downturns resulting from these tensions could materially and adversely affect financial results and future prospects.

These developments have created a dynamic and unpredictable trade landscape, which may adversely impact our business, financial condition, results of operations and prospects.

• The company and nine subsidiaries were added to the U.S. Department of Commerce's Entity List in early 2025.
• Being on the Entity List restricts access to U.S.-origin goods, software, and technology, including certain non-U.S. items with U.S. content.
• The company has implemented export control compliance measures, but there is no guarantee these will be sufficient or strictly followed.
• New legislative efforts, such as the Remote Access Security Act, aim to close gaps that allow PRC-affiliated companies to access restricted technology via cloud services.
• Potential future restrictions on cloud computing leases for AI training could further damage the company's financial condition and operations.

The bill received unanimous bipartisan support in the U.S. House Foreign Affairs Committee.

• Economic sanctions and export controls targeting suppliers or customers can materially damage the company's financial operations.
• New U.S. regulations, including Executive Order 14105 and the America First Investment Policy, restrict investments in Chinese tech firms.
• The company is classified as a 'covered foreign person,' leading to a total prohibition on U.S. persons participating in the Global Offering.
• To comply with these laws, the company has structured its offering to occur exclusively outside the United States in offshore transactions.
• While private equity investment is restricted, a 'publicly traded securities exception' may allow U.S. investors to trade H Shares after the listing.
• The company's limited operating history since 2019 creates additional uncertainty regarding its future performance and growth.

Accordingly, as advised by our International Sanctions Counsel, U.S persons (as defined under the Final Rule) are prohibited to invest in our Company in the Global Offering.

• The company has a limited operating history, making it difficult for investors to assess their ability to market and commercialize solutions.
• Many of the company's products are still in various stages of development rather than being fully realized.
• Predictions regarding future success or viability are inherently less accurate due to the lack of long-term performance data.
• The business faces typical early-stage risks within rapidly evolving and highly competitive industries.
• Historical financial results may not serve as a reliable or meaningful basis for predicting future prospects or financial condition.
• Unforeseen expenses and operational delays could significantly impact the company's ability to achieve its stated goals.

Our historical results may not provide a meaningful basis for evaluating our business, results of operations, financial condition or prospects.

• The company faces significant risks if business strategies based on uncertain future assumptions fail to be implemented effectively.
• Unforeseen external factors, such as regulatory changes or talent shortages, can undermine even well-executed business plans.
• Brand reputation is critical to financial performance, yet marketing efforts like industry conferences carry high costs with no guaranteed return.
• The rapid evolution of AI technologies introduces risks of misuse, bias, and legal challenges that could damage public trust.
• Misuse of AI by third parties or perceived flaws in the technology can lead to litigation and decreased market adoption.
• Strategic initiatives may divert management attention and increase cash outflows, potentially impacting long-term financial stability.

Flaws or misuse of AI technologies may also lead to litigation or other proceedings initiated by certain individuals claiming for infringement of legitimate rights such as privacy or personality rights.

• Misuse or premature application of AI technologies could damage public trust and hinder societal acceptance of AI solutions.
• Inappropriate AI usage, whether intentional or accidental, may lead to negative publicity, legal violations, and regulatory scrutiny.
• The company faces operational risks due to missing title certificates for several leased properties in China, which could lead to eviction.
• Failure to register lease agreements with Chinese authorities may result in administrative fines and legal challenges to property rights.
• The business is susceptible to various legal and administrative proceedings arising from alleged violations of law, contract breaches, or torts.

Any inappropriate, abusive or premature usage of AI technologies, whether actual or perceived, whether intended or inadvertent and whether by us or by third parties, may dissuade prospective customers from adopting AI solutions.

• Legal and administrative proceedings can be costly and disruptive regardless of the actual merit of the claims involved.
• Management attention may be diverted from core business operations to handle litigation, settlements, or injunctive relief.
• The company operates under a complex PRC regulatory scheme involving multiple governmental authorities like the MIIT and CAC.
• Failure to maintain or renew necessary licenses can lead to revenue confiscation, fines, or the total discontinuation of operations.
• Strategic partnerships with global academic institutions are critical for research but remain vulnerable to expansion and maintenance challenges.

If we fail to complete, obtain or maintain any of the required licenses or approvals or make the necessary filings, we may be subject to various penalties, such as confiscation of the revenue that was generated through the implicated operations.

• The company relies on collaborations with academic institutions to recruit talent and build brand recognition, but these partnerships are not guaranteed to continue.
• Potential non-compliance with PRC social insurance and housing provident fund regulations could lead to mandatory supplemental payments and administrative penalties.
• While no material penalties have been issued yet, a change in regulatory interpretation could adversely affect financial conditions and business prospects.
• Technology infrastructure is vulnerable to unexpected failures, natural disasters, human error, and security breaches that could disrupt service availability.
• The company cannot guarantee that existing recovery systems or security protocols will be adequate to prevent data loss or cyberattacks.
• Ongoing reliance on IT networks for R&D and operations creates susceptibility to damage during software upgrades or hardware failures.

We cannot assure you that any applicable recovery system, security protocol, network protection mechanisms or other defense procedures are, or will be, adequate to prevent such network or service interruptions, system failures or data losses.

• Information technology systems are vulnerable to catastrophic events, viruses, and hackers, which could lead to operational disruptions and unauthorized usage.
• Security breaches or system failures risk damaging the company's reputation, reducing customer satisfaction, and incurring significant litigation costs.
• Current insurance coverage is limited, lacking key man life insurance and protection for network infrastructure or property damages.
• The company is actively expanding into international markets, specifically Southeast Asia, which introduces complex regulatory and economic risks.
• Global outreach requires significant management attention and resources to navigate unfamiliar market conditions and intellectual property enforcement.
• Compliance with diverse international laws regarding cybersecurity and data privacy presents a major hurdle for overseas growth.

Any actual or perceived attack or security breach may damage our reputation and brand, expose us to risks of potential litigation and liabilities, and require us to expend significant capital and other resources to alleviate problems caused by such attacks or security breaches.

• The company faces significant challenges from fluctuating exchange rates, trade barriers, and complex geopolitical tensions affecting the AI industry.
• Compliance with international laws is difficult, as meeting the legal requirements of one country may inadvertently violate the regulations of another.
• Natural disasters, force majeure events, and contagious diseases like COVID-19 pose material threats to business continuity and physical infrastructure.
• Vulnerabilities in information systems due to power loss, telecommunications failures, or cyberattacks could lead to severe operational disruptions.
• Intellectual property is critical to the business, yet existing legal protections may not be sufficient to prevent misappropriation by competitors or former employees.
• The company maintains a portfolio of 86 registered patents and 160 copyrights in China but remains at risk regarding the scope of these protections.

In some cases, compliance with the laws and regulations of one country could violate the laws and regulations of another country.

• The patent application process is costly and time-consuming, with no guarantee that proprietary technologies will receive adequate legal protection.
• Failure to identify patentable R&D outputs in a timely manner may allow competitors to commercialize similar solutions without restriction.
• The evolving nature of intellectual property laws in China introduces substantial risks regarding the enforcement and stability of legal rights.
• Litigation to defend intellectual property can lead to significant financial costs, management distraction, and the potential exposure of confidential information during discovery.
• Competitors with greater resources may launch counterclaims or exploit technologies in jurisdictions where the company lacks formal protection.
• Adverse legal rulings could result in existing patents being invalidated, held unenforceable, or interpreted so narrowly that they lose their competitive value.

Furthermore, because of the substantial amount of discovery required in connection with intellectual property litigation, some of our confidential information could be compromised by disclosure during this

• Legal challenges such as counterclaims and third-party administrative actions can lead to the revocation or amendment of existing intellectual property protections.
• Maintaining intellectual property rights requires strict adherence to procedural requirements, fee payments, and documentation deadlines across multiple jurisdictions.
• The loss of patent or copyright protection could allow competitors to enter the market, significantly damaging the company's competitive standing.
• Trade secrets and unpatented know-how are difficult to protect and rely heavily on confidentiality agreements that may be breached intentionally or inadvertently.
• Enforcing trade secret claims is often costly and time-consuming with uncertain legal outcomes.
• The company faces potential litigation from competitors claiming that its solutions infringe upon their own intellectual property rights.

Nevertheless, there can be no guarantee that an employee or a third party will not make an unauthorized use or disclosure of our proprietary confidential information intentionally or inadvertently.

• Competitors may utilize intellectual property litigation as a strategic tool to gain market advantages.
• The hiring of employees from competitors poses a risk of unintentional trade secret misappropriation and subsequent lawsuits.
• Identifying potential infringements is complicated by pending patents and trademarks that are not yet public record.
• IP claims can result in significant financial damages, mandatory product redesigns, or the total prohibition of certain solutions.
• Litigation often diverts management's focus and causes customers to defer or limit their use of affected products.
• New intellectual property acquired by competitors can threaten the viability of products even after they have successfully launched.

We cannot guarantee that such employees will not use their previous employers’ proprietary know-how or trade secrets in their work for us, which could result in litigation against us.

• The company experienced rapid revenue growth between 2022 and 2025, driven primarily by the commercialization of general-purpose large AI models.
• Despite significant revenue increases, there is no guarantee that historical growth rates can be sustained due to market volatility and competition for AI expertise.
• The firm has incurred substantial and increasing net losses, reaching RMB2,357.9 million in the first half of 2025 alone.
• Heavy investments in research and development for foundation models and aggressive sales marketing are the primary drivers of the accumulated losses.
• Future profitability remains uncertain as the company must navigate the dynamic AI market and effectively manage its scaling strategies.
• Failure to address evolving market needs or manage operational expenses could materially damage the company's financial condition.

We may not be able to sustain our historical growth rates, and our historical growth may not be indicative of our future growth or financial results.

• The company anticipates continued significant expenses and net losses for the foreseeable future as it advances its AI performance.
• Revenue generation is heavily dependent on market acceptance of the MaaS platform and the overall development of the AI industry.
• Historical data shows a dramatic escalation in net liabilities, reaching RMB7,080.6 million by mid-2025.
• The company faces substantial liquidity risk due to persistent negative operating cash flows and net current liabilities.
• Profitability remains uncertain, and even if achieved, the company may struggle to sustain it in subsequent periods.
• A significant portion of current liabilities stems from financial instruments issued to investors, which are expected to reclassify to equity upon listing.

Even if we achieve profitability in the future, we may not be able to sustain profitability in subsequent periods thereafter.

• The company may require additional external financing to support expansion if internal cash flow proves insufficient.
• Future operating cash flow is vulnerable to market competition and macroeconomic shifts beyond the company's control.
• Significant credit risk exists due to potential payment delays or defaults from a diverse customer base including enterprises and public entities.
• Trade receivables have shown substantial growth, increasing from RMB27.9 million to over RMB453 million within the tracked period.
• The company faces financial liability risks from redemption rights granted to Pre-IPO investors, which could impact the balance sheet.

We cannot assure you that we will be able to collect all or any of our trade receivables or collect the amount for any unbilled work on time, or at all, after meeting the agreed contractual milestones.

• Financial liabilities are measured at the present value of the highest possible redemption amount, reflecting potential settlement outcomes.
• The company has experienced significant and escalating losses in the carrying amounts of financial instruments issued to investors from 2022 through mid-2025.
• Fluctuations in these carrying amounts pose a material threat to the company's overall financial condition and operational results.
• Substantial capital outlays are required to sustain business operations, R&D strategies, and expansion plans.
• Net cash used in operating activities has increased dramatically, reaching over RMB2.2 billion in 2024.
• The company faces the risk of being unable to secure adequate future financing or only obtaining it on unfavorable terms.

Any significant fluctuations in the changes in the carrying amounts of financial instruments issued to investors may materially affect our financial condition and results of operations.

• The company faces significant capital expenditure requirements and cannot guarantee that future cash flow will be sufficient to meet operational needs.
• Failure to secure additional equity or debt financing could prevent the full implementation of research and development strategies.
• Intangible assets and goodwill are subject to annual impairment tests, which rely heavily on management's subjective judgments and market assumptions.
• Inaccurate estimates regarding revenue growth or discount rates could lead to unexpected impairment losses that adversely affect financial results.
• Share-based payment expenses have increased dramatically, rising from RMB 1.0 million in 2022 to a projected RMB 158.9 million for the first half of 2025.
• The continued use of share incentive plans to attract talent will likely result in the dilution of existing shareholders' equity.

If our estimates and judgments are inaccurate, the recoverable amount determined could be inaccurate and the impairment recognized may not be adequate, and we may need to record additional impairments in the future.

• Adjustments to share-based payment terms and vesting schedules could lead to substantial fluctuations in reporting expenses.
• The company relies on preferential tax treatments for high-tech and small enterprises that are subject to revocation or challenge by PRC authorities.
• Discretionary government grants, which have increased significantly since 2022, are not guaranteed and may be reduced or canceled.
• Operations are heavily influenced by China's shifting economic, political, and social conditions, particularly within the AI industry.
• Future capital raising activities are subject to evolving filing and reporting requirements from the China Securities Regulatory Commission.

If our preferential tax treatments are revoked, become unavailable or if the calculation of our tax liability is successfully challenged by the PRC tax authorities, the discontinuation of any of the various types of preferential tax treatment we enjoy could materially and adversely affect our results of operations.

• The Overseas Listing Trial Measures, effective March 31, 2023, mandate filing requirements for Chinese companies seeking to list securities abroad.
• Companies must file with the CSRC after submitting an offering application, detailing the scope of activities and entities involved.
• Failure to comply with filing procedures or providing false information can result in administrative penalties, including warnings and significant fines.
• Non-compliance may restrict a company's ability to complete its offering, thereby hindering business development and financial prospects.
• The Archives Rules require domestic enterprises and their service providers to establish strict confidentiality and archives management systems.
• These regulations involve multiple Chinese authorities, including the CSRC, the Ministry of Finance, and the National Administration of State Secrets Protection.

If a domestic company fails to complete the filing procedure or conceals any material fact or falsifies any major content in its filing documents, such domestic company may be subject to administrative penalties.

• New Archives Rules and CSRC regulations may impose strict filing requirements that could adversely affect future capital raising activities.
• Failure to obtain necessary regulatory approvals or waivers from Chinese authorities could negatively impact the company's financial condition and share price.
• The majority of the company's assets and senior management are located in China, making service of process and legal enforcement difficult for foreign investors.
• China lacks reciprocal enforcement treaties for civil and commercial judgments with major Western nations including the United States, United Kingdom, and Japan.
• While specific arrangements exist for judgment recognition between Mainland China and Hong Kong, the effectiveness and outcomes of such legal actions remain uncertain.

Currently, China does not have treaties providing for the reciprocal enforcement of judgments in civil and commercial matters by courts with Japan, the United States, the United Kingdom or most other western countries.

• Enforcing foreign legal judgments against the company or its management may be difficult due to lack of reciprocal enforcement treaties.
• Strict PRC regulations on Renminbi conversion and remittance could limit the company's ability to pay dividends or fund offshore capital expenditures.
• Capital account transactions require specific approval or registration with the State Administration of Foreign Exchange (SAFE).
• Fluctuations in exchange rates between the Renminbi and foreign currencies like the U.S. or Hong Kong dollar may diminish asset values and investment returns.
• The company does not currently utilize hedging instruments to mitigate its exposure to foreign currency risk.
• Non-resident investors may be subject to a 10% mainland China withholding tax on dividends and capital gains from share transfers.

As a result, you may encounter difficulty in enforcing foreign judgments against us or our Directors or senior management.

• Foreign individual investors are generally subject to a 20% withholding tax on dividends and capital gains sourced within mainland China.
• A reduced withholding tax rate of 10% typically applies to dividends paid to non-resident individual holders of H Shares based on specific tax arrangements.
• Investors residing in jurisdictions without tax treaties with mainland China must pay the full 20% withholding tax on dividends.
• Gains derived by individuals from the transfer of listed shares may be exempt from individual income tax under specific circulars issued by Chinese authorities.
• Income from the transfer of listed shares on domestic exchanges remains exempt unless the shares are subject to specific sales restrictions.

Non-mainland China resident individual holders who reside in jurisdictions that have not entered into tax treaties with mainland China are subject to a 20% withholding tax on dividends received from us.

• Current Chinese tax law does not explicitly mandate individual income tax on the sale of H shares by non-resident individuals, but future changes could reduce investment value.
• Dividends are strictly paid out of distributable profits calculated under PRC GAAP, which may differ significantly from IFRS standards.
• The company may be unable to pay dividends even during profitable years if there are accumulated losses or mandatory reserve requirements.
• The company's cash flow is dependent on its subsidiaries' ability to distribute profits, which are also subject to PRC legal restrictions.
• Non-compliance with SAFE registration requirements for employee share incentive plans could result in legal sanctions or administrative fines.

As a result, we may not have sufficient, if any, distributable profits to enable us to make dividend distributions to our Shareholders in the future, including periods for which our financial statements indicate that our operations have been profitable.

• PRC residents participating in overseas stock incentive plans must register with the State Administration of Foreign Exchange (SAFE).
• Compliance requires the appointment of domestic qualified agents and overseas-entrusted institutions to manage share transactions.
• Failure to adhere to registration requirements can result in significant fines and legal sanctions for both the company and its employees.
• Regulatory complexities may hinder the company's ability to offer future incentive plans to its directors and staff under PRC law.
• The State Administration of Taxation (SAT) mandates that employees pay individual income tax on exercised options and restricted shares.
• The company bears the legal responsibility to file tax documents and withhold income taxes, facing potential government sanctions for non-compliance.

Failure to complete SAFE registrations may subject them to fines, and legal sanctions.

• The lack of a prior public market for H Shares creates uncertainty regarding future liquidity and trading volume.
• The initial Offer Price is determined by negotiation and may not reflect the actual market value post-offering.
• Stock price volatility may be driven by external factors unrelated to the company's actual operating performance.
• New investors face immediate dilution as the Offer Price exceeds the net tangible book value per share.
• Future capital raises or the exercise of over-allotment options could further dilute existing shareholder interests.
• Dividend payments are not guaranteed and remain subject to board discretion, financial performance, and regulatory constraints.

In addition, the Stock Exchange and other securities markets have, from time to time, experienced significant price and volume volatility that is not related to the operating performance of any particular company.

• The company warns that official government statistics and forecasts regarding the PRC economy may not be entirely accurate or reliable.
• None of the parties involved in the Global Offering have independently verified the data sourced from government publications.
• Investors are cautioned to ignore unauthorized media coverage, press articles, and financial projections not contained within the official prospectus.
• The company disclaims responsibility for any media information that conflicts with the disclosures provided in the formal offering documents.
• A waiver was sought from the Stock Exchange regarding the requirement to have at least two executive directors resident in Hong Kong.
• The management presence waiver is necessary because the group's core business operations are primarily managed and conducted within the PRC.

Certain facts, forecasts and other statistics obtained from government publications contained in this prospectus may not be reliable in terms of accuracy, competence or reliance.

• The company has appointed Dr. Liu Debing and Mr. Cheng Ching Kit as authorized representatives to serve as the primary communication channel with the Stock Exchange.
• All directors, including those not resident in Hong Kong, must maintain constant accessibility and possess valid travel documents for potential meetings with regulators.
• Maxa Capital Limited has been appointed as a Compliance Advisor to provide an additional layer of communication between the company and the Stock Exchange.
• The text outlines strict protocols for updating the Stock Exchange regarding any changes to key personnel or contact information.
• A waiver is sought regarding Rule 10.04 to allow close associates of existing shareholders to participate as cornerstone investors in the offering.
• The company must satisfy specific conditions under the Listing Rules to permit share allocations to existing shareholders or their nominees.

In the event that a Director expects to travel, he/she will endeavor to provide the phone number of the place of his/her accommodation to our Authorized Representatives or maintain an open line of communication via his/her mobile phone.

• The Listing Rules permit existing shareholders of Specialist Technology Companies to participate in IPOs to meet funding needs.
• Shareholders with a 10% or greater stake must subscribe as cornerstone investors, while those with less than 10% may act as cornerstone investors or placees.
• Strict anti-preference rules ensure that existing shareholders receive no special allocation advantages over other investors, aside from assured entitlement at the IPO price.
• The document identifies specific cornerstone investors including Qizhi SP, Structured Credit SP Fund, and Luster LightTech International.
• Complex ownership structures reveal significant state involvement, with several investment entities ultimately controlled by the Beijing SASAC.

In the case of subscription as a placee, the applicant and its sponsors must confirm that no preference in allocation was given to the existing shareholder.

• The text details complex ownership structures involving various Beijing-based state-owned assets supervision and administration commissions (SASACs).
• Multiple investment funds, including the AI Fund and Daxing Industrial Fund, are identified as close associates due to their ultimate control by PRC governmental bodies.
• Luster LightTech International is established as a close associate of its parent company, which was already a shareholder prior to the global offering.
• The Structured Credit SP Fund is linked to the Tsinghua University Education Foundation, which supports academic standards and research facilities in China.
• Tsinghua University is identified as a beneficial owner of both the Structured Credit SP Fund and Tsinghua Technology, making them close associates.
• The company has applied for specific waivers from the Stock Exchange to allow these associated entities to participate as cornerstone investors in the Global Offering.

The objects of the Tsinghua University Education Foundation include improving quality of education and academic standards in China.

• The Company and Sole Sponsor confirm that no preferential treatment was given to specific cornerstone investors like Qizhi SP and Luster LightTech beyond assured entitlement at the Offer Price.
• Regulatory guidelines generally prohibit share allocations to 'connected clients' of syndicate members without prior written consent from the Stock Exchange.
• CICC Financial Trading Limited (CICC FT) is identified as a connected client of the Sole Sponsor, CICCHKS, as they belong to the same corporate group.
• CICC FT will utilize cross-border delta-one OTC swap transactions to pass the economic risks and returns of the Offer Shares to independent third-party clients.
• The Stock Exchange granted a waiver for CICC FT's participation provided the terms are no more favorable than those offered to other cornerstone investors.

CICC FT will hold the Offer Shares on a non-discretionary basis to hedge the Gaoyi OTC Swaps, while the economic risks and returns of the underlying Offer Shares are passed to the CICC FT Ultimate Clients (Gaoyi).

• CICC FT confirms it will not receive preferential treatment in share allocation beyond its cornerstone investment agreement.
• The company and its coordinators have provided all necessary written confirmations to the Stock Exchange per regulatory guidelines.
• Directors accept full collective and individual responsibility for the accuracy and completeness of the prospectus information.
• The prospectus confirms that no material omissions exist that would render the document's statements misleading or deceptive.
• The CSRC filing process for the H Shares listing and Global Offering was completed between July and December 2025.

Our Directors, having made all reasonable enquiries, confirm that, to the best of their knowledge and belief, the information contained in this prospectus is accurate and complete in all material respects and not misleading or deceptive.

• The Hong Kong Offer Shares are offered exclusively based on the information and representations explicitly contained within the prospectus.
• No unauthorized person or entity is permitted to provide information or make representations regarding the Global Offering outside of this document.
• The offering is fully underwritten by designated Hong Kong and International Underwriters, with the latter agreement expected by early January 2026.
• The prospectus does not guarantee that the company's affairs will remain unchanged after the date of publication or that the information remains current indefinitely.
• Public offering and distribution of the prospectus are strictly limited to Hong Kong, with specific prohibitions against sales in the PRC.
• Applicants are legally deemed to represent that they are not associates of the company's directors or existing shareholders.

No action has been taken to permit a public offering of the Offer Shares or the general distribution of this prospectus in any jurisdiction other than in Hong Kong.

• The company has applied for a listing of H Shares on the Hong Kong Stock Exchange, including shares from the Global Offering and converted unlisted shares.
• Trading is expected to commence at 9:00 a.m. on January 8, 2026, under the stock code 2513, with shares traded in board lots of 100.
• Under local ordinances, any share allotment becomes invalid if the Stock Exchange refuses the listing within a specific three-to-six-week window.
• The H Shares will be eligible for deposit, clearance, and settlement through the Central Clearing and Settlement System (CCASS).
• The company explicitly disclaims responsibility for any tax liabilities incurred by investors, recommending consultation with professional advisors.

Under section 44B(1) of the Companies (Winding Up and Miscellaneous Provisions) Ordinance, any allotment made in respect of any application will be invalid if the listing of, and permission to deal in, the H Shares on the Stock Exchange is refused.

• H Shares will be registered in Hong Kong through Tricor Investor Services Limited, while the principal register remains at the PRC head office.
• Transactions involving H Shares are subject to Hong Kong stamp duty, and dividend payments are dispatched via ordinary post at the shareholder's risk.
• Financial data uses specific exchange rate conversions for RMB, USD, and HKD, though these rates are for convenience and do not guarantee actual conversion liquidity.
• The English version of the prospectus takes legal precedence over Chinese translations, except for the names of PRC entities where the original Chinese names prevail.
• Rounding adjustments in financial tables may result in arithmetic discrepancies between individual figures and the listed totals.
• The document identifies key executive and non-executive directors, including their residential addresses and nationalities within the PRC.

No representation is made that the amounts denominated in one currency could actually be converted into the amounts denominated in another currency at the rates indicated or at all.

• The document lists key administrative personnel including Dr. Xie Deren and Mr. Tang Ying, primarily based in Beijing and Hong Kong.
• A notable governance detail specifies that Supervisor Mr. Pei Bo will resign upon the completion of the company's listing per PRC regulations.
• China International Capital Corporation (CICC) serves as the Sole Sponsor and a primary Overall Coordinator for the global offering.
• A large syndicate of financial institutions, including Huatai Financial and BOCOM International, is managing the bookrunning and lead management roles.
• The legal framework for the offering involves multiple jurisdictions, with Linklaters handling Hong Kong and U.S. laws while Tian Yuan Law Firm manages PRC legal matters.
• Specialized legal counsel from Bayfront Law LLC and King & Wood Mallesons has been retained for Singaporean data protection and international sanctions compliance.

In accordance with the applicable PRC laws and regulation, the Supervisor will resign as Supervisor of the Company with effect from the completion of Listing.

• The document lists the extensive legal and financial advisory network supporting the offering, including Baker & McKenzie and KPMG.
• Corporate governance is established through specialized committees focusing on audits, remuneration, nominations, and ESG strategy.
• The company maintains a dual presence with headquarters in Beijing's Haidian District and a principal place of business in Wanchai, Hong Kong.
• Industry data provided by Frost & Sullivan serves as the foundation for the market analysis, though government statistics remain unverified by the underwriters.
• The text transitions into a broader industry overview, positioning Artificial Intelligence as the catalyst for the Fourth Industrial Revolution's inflection point.

The Fourth Industrial Revolution represents a profound global technological transformation, driven by the convergence of digital, physical, and biological systems.

• The Fourth Industrial Revolution is defined by the convergence of physical, digital, and biological worlds, moving at a faster pace than previous revolutions.
• Artificial Intelligence serves as the primary driver of this era, enabling intelligent automation and the merging of physical and digital realms.
• AI development relies on three core pillars: exponential computing power, innovative algorithms like Transformer architectures, and massive volumes of high-quality data.
• Global economic impact is projected to reach over US$20 trillion by 2030, with AI influencing 20% of all business decision-making.
• China's AI market is experiencing rapid acceleration, with a projected compound annual growth rate of 35.5% leading into 2030.

Unlike the first three industrial revolutions which were powered by steam, electricity and computers and Internet, the Fourth Industrial Revolution is unfolding at a faster pace, on a broader scale, and with deeper impacts on society and the economy.

• AI technology is transitioning from discriminative models focused on recognition to Large Language Models (LLMs) capable of content generation.
• LLMs utilize Transformer architectures with billions of parameters, allowing them to understand context and generate coherent text, code, and media.
• Unlike traditional AI that requires high-quality labeled data, LLMs rely on massive amounts of unstructured data during their pre-training phase.
• Discriminative AI is limited by its task-specific design, struggling with deep semantic comprehension and multi-turn dialogues.
• LLMs possess a degree of autonomy and decision-making capability that enables them to act as independent agents in response to external changes.
• The hardware requirements for LLMs are significantly higher, necessitating massive distributed computing clusters and high-performance GPUs.

With autonomy, perception, decision-making and execution capabilities, LLMs can act independently in response to external changes and set goals.

• AI is transitioning from task-specific Artificial Narrow Intelligence (ANI) to Artificial General Intelligence (AGI) which matches human cognitive capabilities.
• Large Language Models (LLMs) are the primary drivers of this shift due to breakthroughs in parameter scale, multimodal fusion, and self-evolution.
• The development of LLMs is categorized into five stages: pre-training, alignment and reasoning, self-learning, self-perception, and consciousness.
• The industry's developmental framework is modeled after cognitive science, moving from basic linguistic skills to abstract thinking and self-awareness.
• Leading institutions like OpenAI utilize similar five-level frameworks to track the progression from conversational agents to autonomous and organizational AI.

The sequence is closely modeled on developmental stages observed in cognitive science, where intelligence is seen as progressing from basic perception and linguistic capability, through the acquisition of reasoning and problem-solving skills, to the emergence of self-awareness and abstract thinking.

• The commercialization of Large Language Models (LLMs) in China is primarily driven by institutional and enterprise demand rather than individual consumers.
• Enterprises demonstrate a significantly higher willingness to pay for LLM integration compared to the consumer market, where monetization remains in early stages.
• Key enterprise applications include automated customer service, personalized marketing content generation, and productivity tools for workflow automation.
• The Chinese LLM market is projected to grow from RMB5.3 billion in 2024 to RMB101.1 billion by 2030, maintaining a robust CAGR of 63.5%.
• Institutional customers are expected to contribute the vast majority of market revenue, estimated at RMB90.4 billion of the total market by 2030.
• Business models are diversifying into cloud-based and on-premise deployment strategies to meet varying corporate security and infrastructure needs.

In contrast, the enterprise scenario is the main driver of growth in LLM market in China.

• The Chinese enterprise LLM market is bifurcated into cloud-based and on-premise deployment models, each catering to different business scales and security needs.
• On-premise deployment currently dominates the market revenue, favored by large enterprises for its superior data control and customization capabilities.
• The total enterprise LLM market in China is projected to explode from RMB4.7 billion in 2024 to RMB90.4 billion by 2030.
• While on-premise solutions hold the largest market share, cloud-based deployments are expected to grow at a faster CAGR of 69.4% through 2030.
• The industry value chain distinguishes between 'independent' LLM providers, who are natively built around the technology, and 'non-independent' providers who leverage existing product portfolios.

This setup allows organizations to build, train and use LLMs tailored to their needs, all within their own infrastructure.

• Cloud-based deployment offers enterprises 'ready-to-use' and scalable LLM capabilities through standardized APIs.
• Software and device manufacturers can efficiently embed text, image, and voice generation into products like AI glasses and robots.
• The cloud model eliminates the need for intensive upfront investments in computing resources and complex engineering.
• Pay-as-you-go pricing models transform AI from an exclusive tool for large corporations into universal infrastructure for SMEs.
• Agile cloud interfaces allow enterprises to integrate models within hours, significantly shortening the cycle from planning to commercial deployment.

The inclusive nature of cloud-based deployment helps transform AI technology from a tool exclusive to large enterprises into a universal infrastructure for all types of enterprises, accelerating the intelligent transformation of the entire industry.

• Cloud-based LLMs offer rapid iteration and access to advanced capabilities through continuous fine-tuning by expert teams.
• Market concentration is increasing as leading cloud providers leverage massive computational resources and ecosystem advantages to outperform smaller players.
• The future of cloud AI is shifting toward multimodal integration, processing text, visual, audio, and motion data for holistic decision-making.
• On-premise deployment is becoming the preferred strategy for sectors like finance and government that prioritize data sovereignty and security.
• Hosting models locally allows for deep industry-specific customization using proprietary data that cannot be shared with external cloud providers.
• On-premise solutions reduce dependency on external networks, ensuring system reliability and responsiveness for mission-critical operations.

In the future, LLMs will expand from text processing to image recognition, video analysis, 3D modeling and more, enabling holistic AI solutions and better decision-making for enterprises.

• The AI market is evolving into a dual-engine model where open-source and closed-source systems coexist to balance flexibility with professional stability.
• Open-source models drive rapid technology diffusion and lower innovation costs, while closed-source models lead in high-performance commercial applications.
• A maturing value chain in China—spanning data collection to industry-specific applications—is fostering a collaborative and sustainable market growth.
• The market structure is bifurcating into general-purpose digital infrastructure and specialized vertical ecosystems for sectors like healthcare and finance.
• The competitive landscape distinguishes between 'independent' providers focused solely on LLMs and 'non-independent' providers leveraging existing product portfolios.
• Market data from 2024 identifies 'The Company' as the largest independent LLM provider in China and the second-largest provider overall by revenue.

Independent providers are characterized by being natively built around LLM technology, products and business models from the early stages of their operations.

• Company A focuses on intelligent audio technology and smart city solutions, operating primarily out of Hefei since 1999.
• Company B is a massive e-commerce and cloud conglomerate with nearly 200,000 employees and listings on both the HKEX and NYSE.
• Company C specializes in computer vision and automotive solutions, maintaining a smaller workforce of under 5,000 employees.
• Company D provides a diverse portfolio including intelligent driving and cloud services, with a significant workforce of nearly 400,000.
• The text distinguishes Model-as-a-Service (MaaS) from traditional Platform-as-a-Service (PaaS) by its focus on pre-trained large models.
• MaaS operates at a higher level of abstraction than PaaS, targeting both enterprises and developers rather than just development teams.

The MaaS platform provided by the Company, on the other hand, operates at a higher level of abstraction, focusing specifically on delivering pre-trained large models and model development toolkits as services.

• Cloud platforms are evolving to provide AI-native development services that abstract away underlying infrastructure management.
• Leading vendors maintain a competitive edge by offering a multi-layered model matrix ranging from lightweight edge models to massive flagship models.
• A positive feedback loop is created by using real-world client data to iteratively lower training costs while improving model performance.
• Success depends on flexible business models including on-demand, subscription, and one-time deployment options to suit diverse enterprise budgets.
• The creation of a 'closed-loop' ecosystem involving chip manufacturers, developers, and industry-specific clients forms a significant barrier to entry.
• Open-sourcing models has become a strategic tool for vendors to accumulate a vast developer base and extend technological reach.

This forms a positive feedback loop featuring low training costs, strong model performance and high adaptability to applications.

• The 'feedback' ecosystem model creates a competitive barrier by integrating open-source communities with hardware-software synergy.
• China's Large Language Model (LLM) market faces a significant talent barrier due to the high demand for experts with deep technical backgrounds.
• The industry analysis is supported by a commissioned Frost & Sullivan report costing RMB600,000, independent of the listing's success.
• Data and forecasts are derived from a mix of in-house databases, government publications, and independent third-party reports.
• The regulatory landscape for AI in China is governed by specific provisions on deep synthesis and internet-based information services.
• Compliance is overseen by multiple authorities including the CAC, the MIIT, and the Ministry of Public Security.

The 'feedback' ecosystem model... achieves a deep integration of resources, capabilities and clients, forming an unique ecosystem barrier that is difficult to replicate.

• The Deep Synthesis Provisions mandate that service providers verify user identities and label AI-generated content to prevent the spread of illegal information.
• Service providers with social mobilization capabilities must undergo security assessments and comply with specific filing requirements for recommendation algorithms.
• The Interim Measures on Generative AI hold service providers legally responsible as the 'producers' of any online information generated by their systems.
• Providers are required to implement content filtering, promptly remove illegal material, and optimize models to prevent the recurrence of prohibited outputs.
• Strict data privacy rules prohibit the collection of unnecessary personal information and the unlawful retention of user input records.
• Regulations specifically demand that providers prevent minors from developing addiction to or excessive dependence on generative AI services.

Service providers shall be responsible as producers of the online information.

• Service providers must ensure training data and base models originate from lawful sources and respect intellectual property rights.
• Providers are required to implement measures ensuring the authenticity, accuracy, objectivity, and diversity of training data.
• AI services with public opinion or social mobilization capabilities must undergo security assessments and formal filing with the CAC.
• Compliance involves a multi-stage content review process, including sensitive vocabulary databases and manual re-examination of high-risk content.
• New regulations mandate the identification and labeling of AI-generated and synthesized content to ensure transparency.

We utilized our models to improve contextual comprehension, enabling detection of illegal information.

• New regulations mandate that all AI-generated content must feature both explicit visual indicators and implicit digital watermarks.
• Service providers and platforms are legally responsible for ensuring that AI-generated text, audio, and video are clearly identified to users.
• The company has completed all necessary algorithm and generative AI filings required by PRC legal authorities.
• Directors and legal advisors confirm the company has faced no material penalties or investigations regarding AI-related laws during the track record period.
• National industrial catalogs currently classify big data, cloud computing, and blockchain as 'encouraged' categories for development.
• Unauthorized intrusion into strategic computer systems or the dissemination of destructive viruses remains a criminal offense under internet security laws.

Explicit markings, such as visible text prompts and visual indicators, must be presented prominently to users.

• Internet service providers are legally prohibited from leaking state secrets, spreading false commercial information, or infringing on intellectual property.
• Providers must maintain user logs, including login and exit times, for a minimum of 60 days to facilitate the detection of illegal information.
• The Cybersecurity Law mandates a graded protection system to prevent data disclosure, theft, or tampering while requiring emergency response plans.
• Network operators are required to provide technical assistance to national security authorities for criminal investigations and state security protection.
• National information security follows a principle of 'independent grading,' where entities determine their own security levels based on specific regulatory guidelines.
• The State Security Law emphasizes achieving 'secure and controllable mastery' of core technologies and critical infrastructure to safeguard cyberspace sovereignty.

The government shall intensify network governance, prevent, suppress and lawfully punish cybercrimes including cyberattacks, network intrusions, cyber espionage and dissemination of illegal or harmful information, thereby safeguarding national cyberspace sovereignty, security and

• The Criminal Law of the PRC imposes criminal liability on network service providers who fail to fulfill security management obligations after being ordered to rectify.
• The Data Security Law establishes a classification and grading protection system to safeguard individual and organizational rights while promoting the digital economy.
• National security reviews are conducted for data processing activities that may impact the state, with final decisions being legally binding.
• Organizations must implement whole-process data security management systems, including technical safeguards and employee training.
• Critical Information Infrastructure (CII) is defined as systems in vital sectors like energy, finance, and national defense that could endanger the public if compromised.
• Competent authorities are responsible for identifying and notifying specific operators designated as part of the Critical Information Infrastructure.

Decisions rendered in accordance with the law through such reviews shall be final.

• The Cybersecurity Review Measures mandate national security reviews for critical information infrastructure operators (CIIOs) and network platform operators.
• Network Platform Operators with over one million users must undergo a cybersecurity review before seeking a listing in a foreign country.
• The Security Assessment Measures establish strict thresholds for outbound data transfers, including specific limits on sensitive personal information.
• Industrial and telecommunication data processors are now required to implement rigorous data classification and grading systems.
• New administrative measures impose comprehensive obligations on data handling, from collection and storage to destruction and emergency planning.

In addition, Network Platform Operators holding personal information of more than one million users that seek listing in a foreign country are obliged to apply for a cybersecurity review by the Cybersecurity Review Office.

• The Overseas Listing Trial Measures mandate that domestic companies comply with national security and data protection laws before listing on foreign exchanges.
• Companies must complete required national security reviews before submitting listing applications to overseas regulatory agencies or trading venues.
• New provisions from March 2024 offer exemptions for data security assessments if a non-CIIO processor transfers personal data of fewer than 100,000 individuals annually.
• Specific thresholds exist for mandatory standard contracts or certifications, triggered when transferring data for 100,000 to 1,000,000 individuals.
• Data processors are exempt from security assessments for outbound transfers if the data has not been officially designated as 'important data' by the state.
• The regulatory landscape continues to evolve with the introduction of the Cyber Data Security Regulations in late 2024.

If the proposed overseas offering and listing necessitates a national security review, relevant security review procedures shall be completed according to law before the application for such offering and listing is submitted to any overseas parties.

• The Cyber Data Security Regulations effective in 2025 consolidate requirements from existing cybersecurity and personal information laws.
• Data processors are mandated to implement technical safeguards like encryption and access control to prevent unauthorized data leaks or tampering.
• National security reviews are required for data processing activities that could impact state security, aligning with cross-border transfer protocols.
• The PRC Civil Code and Personal Information Protection Law establish core principles of legitimacy, necessity, and minimal scope for data handling.
• Specific protections are mandated for minors, requiring online service providers to manage their access duration and consumption habits.
• Important data must be officially designated by authorities before specific outbound transfer security assessments are triggered.

Notably, if the data has not been notified or publicly released as important data by relevant regions or departments, there is no need to apply for an outbound important data transfer security assessment before transferring the data overseas.

• Personal information handlers must implement security measures or face severe penalties including service termination and confiscation of illegal income.
• Internet information service providers are required to establish real-name authentication and maintain robust systems for ecological governance and emergency response.
• Service providers must maintain visible portals for public complaints and whistleblowing to ensure timely handling of grievances.
• Entities processing data for over ten million individuals must undergo mandatory compliance audits at least once every two years.
• The CAC has the authority to mandate third-party professional audits of personal information processing activities under specific circumstances.
• Mobile application providers are strictly prohibited from utilizing software to threaten national security or infringe upon the rights of minors.

The internet information service providers shall set up a convenient portal for complaints and whistleblowing at an eye-catching position and timely handle the complaints and whistleblowing of users and the public.

• Mobile application providers must verify user identities and implement management systems to ensure content security and user rights protection.
• Regulatory bodies prohibit the collection of personal information irrelevant to services and mandate clear, voluntary user consent without coercive methods.
• App operators are forbidden from using default settings, bundling clauses, or making consent a prerequisite for accessing basic services.
• The MIIT conducts inspections to target deceptive practices, such as misleading users into downloading apps or excessively requesting permissions.
• New mandates require app providers to maintain a list of collected personal information and complete formal filing procedures with provincial authorities.

The data processors shall clearly outline the purpose, method, and scope of sensitive personal information processing activities, maintain a list of collected personal information, and shall not use methods such as default ticking, reducing text size, or lengthy texts to induce users’ consent.

• The 2019 Foreign Investment Law established a unified legal foundation for foreign entities, replacing several legacy joint-venture laws.
• China employs a 'Negative List' system where foreign investment is prohibited or restricted in specific industries but permitted elsewhere under national treatment.
• The Reporting Measures require foreign investors to submit comprehensive data, including initial, change, and annual reports, to commerce departments.
• A security review mechanism is mandatory for investments in national defense or major sectors where foreign investors acquire actual control.
• Major sectors subject to security review include critical infrastructure, energy, information technology, and essential cultural products.
• The 2024 Negative List serves as the current regulatory benchmark for determining which industries remain restricted to foreign capital.

The following categories of foreign investment shall require foreign investors or domestic relevant parties to proactively declare the security review prior to implementation: (i) investments in national defense security-related sectors... and (ii) investments in major sectors affecting national security.

• Telecommunications services in China are strictly categorized into infrastructure and value-added services, both requiring specific operating licenses from the MIIT.
• Internet information services are classified as value-added telecommunications services, necessitating compliance with rigorous administrative measures and catalogs.
• Foreign equity ownership in value-added telecommunications enterprises is generally capped at 50%, with specific exceptions for sectors like e-commerce and call centers.
• The 2024 Negative List explicitly restricts foreign investment in companies providing internet content through mobile apps and websites.
• Domestic license holders are strictly prohibited from leasing, selling, or transferring telecommunications licenses to foreign investors or providing them with operational resources.
• Regulations mandate that foreign investors must establish specific foreign-invested enterprises to apply for the necessary telecommunications business operation licenses.

A PRC domestic company that holds a value-added telecommunications business operation licenses is prohibited from leasing, transferring or selling such license to foreign investors in any means.

• Commercial internet content providers must obtain a specific value-added telecommunications business license from Chinese authorities.
• Patent law categorizes protections into invention, design, and utility models with varying durations from 10 to 20 years.
• Trademark registrations last for ten years and require formal filing of license agreements with the trademark bureau to remain valid.
• Copyright protection applies automatically to works of literature, science, and software, granting owners both personal and property rights.
• The National Copyright Administration and China Copyright Protection Center oversee the formal registration and management of computer software copyrights.
• Domain name services are supervised by the MIIT, which requires formal registration procedures for an applicant to become a holder.

Implementation of a patent without the authorization of the patent holder shall constitute an infringement of patent rights, and shall be held liable for compensation to the patent holder and may be imposed a fine, or even subject to criminal liabilities.

• Internet information service providers must legally own and register their domain names under the entity or its senior management.
• The PRC Labor Law and Labor Contract Law mandate written employment contracts and impose strict requirements on hiring and dismissal.
• Employers are legally obligated to ensure employee rights to rest and provide wages that meet or exceed local minimum standards.
• Companies must establish labor safety and sanitation systems that strictly adhere to state standards and provide employee education.
• The Social Insurance Law requires enterprises to provide comprehensive welfare schemes including pension, medical, and unemployment insurance.
• Failure to make full and timely social insurance contributions can result in late fees, administrative fines, or criminal liabilities.

Violations of the Labor Contract Law of the PRC and the Labor Law of the PRC may result in the imposition of fines and other administrative liabilities and/or incur criminal liabilities in the case of serious violations.

• New entities must register for the housing accumulation fund and open employee accounts within 30 days of establishment or recruitment.
• Failure to comply with housing fund registration can result in fines up to RMB50,000 and potential compulsory enforcement by the People’s Court.
• The Overseas Listing Trial Measures establish a strict filing and regulatory regime for PRC domestic enterprises seeking international listings.
• Non-compliance with overseas listing procedures can lead to corporate fines up to RMB10,000,000 and personal fines for responsible officers.
• Listed issuers are required to report material events, such as changes in control or regulatory investigations, to the CSRC within three working days.
• New provisions strengthen confidentiality and archives administration for domestic enterprises involved in overseas securities offerings.

Any entity failing to complete the relevant procedure within the time limit will be fined RMB10,000 to RMB50,000.

• Strict compliance procedures are required for overseas listed entities when disclosing information involving state secrets or national security interests.
• The 'Full Circulation' policy allows domestic unlisted shares of H-share companies to be listed and traded on the Stock Exchange.
• Shareholders have the flexibility to determine the volume of shares for circulation through mutual consultation and regulatory compliance.
• Once domestic unlisted shares are converted and listed on an overseas exchange, they cannot be converted back to the domestic market.
• The CSRC and Shenzhen Stock Exchange have established specific implementation measures for cross-border share transfers and settlement management.
• Corporate entities are subject to the PRC Enterprise Income Tax Law, which governs the taxation of income for domestic and overseas-listed firms.

After domestic unlisted shares are listed and circulated on the Stock Exchange, they may not be converted back domestically.

• Enterprises in China are classified as resident or non-resident based on their legal establishment and actual place of management.
• Resident enterprises are generally subject to a uniform 25% income tax rate on their worldwide income, with reductions for small low-profit or high-tech firms.
• Dividend distributions to Hong Kong entities can benefit from a reduced 5% withholding tax rate if specific ownership thresholds are met.
• Non-resident taxpayers utilize a self-assessment system for treaty benefits, requiring them to retain documentation for potential tax authority inspections.
• Value-added Tax (VAT) applies to most commercial activities in China, with a standard 6% rate for services and intangible assets.

Resident enterprises refer to enterprises that are established in accordance with the law within the territory of China, or enterprises that are established in accordance with foreign (regional) laws but have their actual place of management within the territory of China.

• China implemented a series of VAT reforms between 2017 and 2019, progressively lowering the top tax rates from 17% to 13%.
• The Renminbi is freely convertible for current account items like trade and dividends but remains restricted for capital account items.
• Domestic companies must register overseas listings with the State Administration of Foreign Exchange (SAFE) within 15 working days.
• Funds raised through overseas listings must be used in a manner consistent with public disclosure documents and are generally expected to be repatriated.
• Recent administrative shifts allow qualified banks to handle foreign exchange registrations directly under SAFE supervision to simplify the process.

Renminbi is freely convertible for payments of current account items, such as trade and service-related foreign exchange transactions and dividend payments, but not freely convertible for capital account items.

• SAFE Circular 16 establishes a 100% tentative percentage for foreign exchange settlement based on actual business needs of domestic institutions.
• Enterprises are strictly prohibited from using settled foreign exchange for securities investment, inter-enterprise borrowings, or non-self-use real estate.
• Outbound profit remittance requires rigorous verification of board resolutions, tax records, and audited financial statements to ensure transaction genuineness.
• Foreign-invested enterprises (FIEs) are permitted to use converted Renminbi for domestic equity investments provided they comply with the negative list.
• Recent reforms allow qualified enterprises to make domestic payments from capital accounts without providing prior proof of veracity for every transaction.

The tentative percentage of foreign exchange settlement for foreign currency earnings in capital account of domestic institutions is 100%, subject to adjustment by SAFE in due time in accordance with international revenue and expenditure situations.

• PRC laws mandate that foreign-invested enterprises pay dividends only from accumulated profits after setting aside 10% for general reserves.
• Companies in China must continue reserve allocations until they reach 50% of registered capital and must offset prior losses before distribution.
• The PRC Anti-Unfair Competition Law requires operators to adhere to principles of integrity and business ethics to protect market order.
• New interim provisions specifically target unfair competition in cyberspace to regulate the digital economy and encourage innovation.
• U.S. export controls, managed by the BIS under the EAR, regulate the transfer of commercial, dual-use, and military-related technologies.

A PRC company, including foreign-invested enterprise, is required to set aside as general reserves at least 10% of its after-tax profit, until the cumulative amount of such reserves reaches 50% of its registered capital.

• The Bureau of Industry and Security (BIS) regulates all U.S.-origin items globally, including those in transit through the United States.
• Foreign-made items are subject to the EAR if they contain more than a de minimis amount of controlled U.S. content, typically 25% or 10% for sanctioned nations.
• The 'direct product' rule extends U.S. jurisdiction to foreign items created using specific highly-controlled U.S. software or technology.
• Violations of export controls carry severe penalties, including fines up to $374,474 or twice the transaction value, and up to 20 years in prison.
• Determining license requirements involves checking the Commerce Control List (CCL), country-based restrictions, and restricted party lists.

Violations of U.S. export controls may have serious consequences including, but not limited to, civil monetary penalties of up to $374,474 in 2025 or twice the value of the transaction, whichever is greater; or criminal penalties of up to $1 million per violation and/or up to 20 years in prison.

• The Bureau of Industry and Security (BIS) added the company and nine subsidiaries to the Entity List on January 16, 2025, restricting access to items subject to the EAR.
• Legal guidance clarifies that Entity List restrictions do not automatically apply to legally distinct parent companies, subsidiaries, or sister companies unless specifically listed.
• The Group has implemented comprehensive export control compliance measures across all entities to mitigate risks following the Entity List Addition.
• Suppliers and buyers are strictly liable for unauthorized exports, reexports, or transfers of EAR-governed items to listed entities.
• Liability under the EAR extends beyond direct violations to include aiding, abetting, conspiring, or acting with 'knowledge' of a violation.
• The EAR defines 'knowledge' broadly to include the conscious disregard of facts or the willful avoidance of information regarding potential violations.

The EAR defines “knowledge” as including “positive knowledge that the circumstance exists or is substantially certain to occur,” as well as “an awareness of a high probability of its existence or future occurrence,” which is “inferred from evidence of the conscious disregard of facts known to a person and is also inferred from a person’s willful avoidance of facts.”

• Founded in 2019 by a team from Tsinghua University, the company is a leading Chinese AI firm focused on achieving artificial general intelligence (AGI).
• The company transitioned from foundational pre-training frameworks in 2020 to releasing massive models like GLM-130B and ChatGLM by 2023.
• Zhipu QingYan was among the first large model products to successfully complete regulatory filing in China, marking a significant compliance milestone.
• The 2024-2025 roadmap shows a rapid expansion into multimodal capabilities, including visual comprehension, video generation, and emotional voice models.
• The company is pioneering 'AutoGLM' agents capable of autonomous device operation and multi-step reasoning through 'rumination' and deep research features.

Our Company draws on the academic rigor and innovative spirit of its founding team to advance cutting-edge developments in AI.

• The company originated with a diverse ownership structure involving individual founders, institutional investors like CAS Star, and the Beijing Lianpai holding entity.
• In 2021, the company established employee ownership platforms, Huihui and Zhideng, through capital transfers from Beijing Lianpai to incentivize staff.
• A significant transition occurred in March 2025 when the company converted from a limited liability company into a joint stock company based on audited net assets.
• Pre-IPO strategic growth was supported by multiple funding rounds, including a Series Angel Financing involving the transfer of intellectual property rights valued at over RMB20 million.
• A one-for-ten share subdivision was scheduled for mid-2025 to adjust the nominal value of shares immediately prior to the public listing.
• Legal advisors have confirmed that all capital transfers, share issuances, and subdivisions comply with relevant PRC laws and SAMR registrations.

Innovation Zhiyuan subscribed for registered capital of our Company in the amount of RMB526,316 by transferring intellectual property rights with an appraised value of approximately RMB20.37 million.

• The company successfully completed its Series A financing in May 2022, raising a total of RMB 152 million from eight different investment entities.
• Series B1 financing followed in February 2023, securing RMB 208 million primarily from Junlian Xiangdao and Qiming entities.
• The Series B2 financing round was executed in two stages during mid-2023, totaling over RMB 410 million in consideration.
• Tianjin Sankuai Technology emerged as a major investor during the first stage of Series B2, contributing RMB 300 million alone.
• The financing trajectory shows rapid scaling, with the company moving from Series A to Series B3 within a span of less than two years.
• Several capital transfers between shareholders occurred alongside these rounds, conducted through arm's length negotiations without increasing registered capital.

Such capital transfers were conducted based on arm’s length negotiation between the Shareholders and did not involve any increase of registered capital by our Company.

• The company executed a multi-stage financing strategy between November 2023 and January 2024, raising over RMB 608 million from investors like Trend Mega and Shanghai Yunya.
• Series B4 financing, completed in August 2024, significantly increased the capital base with a total consideration of approximately RMB 1.23 billion.
• High-profile institutional investors participated in the Series B4 round, including Tencent Investment, TAL Education, and the Social Security Zhongguancun Innovation Fund.
• Internal capital transfers between shareholders, such as those involving CAS Star and Shanghai Yunya, were conducted via arm’s length negotiations without increasing the company's total registered capital.
• The company continued its aggressive expansion of registered capital into late 2024 through the Series B5 financing rounds, involving entities like Shanghai Feiya.

Such capital transfer was conducted based on arm’s length negotiation between the Shareholders and did not involve any increase of registered capital by our Company.

• The text details a complex series of capital injections through Series B5 and B6 financing rounds involving numerous institutional investors.
• Major subscribers include specialized funds like the AI Fund, Prosperity7, and the Social Security Zhongguancun Innovation Fund.
• The Series B6-2 financing represents a significant single investment from the Haihe Fuxin Youda Fund totaling 950 million RMB.
• Internal capital transfers occurred in late 2024, involving the movement of registered capital between entities like Huapin Borui and Xiaofeng Technology.
• The financing timeline extends from late 2024 into May 2025, demonstrating a rapid and continuous scaling of the company's capital base.
• Total considerations for the listed sub-rounds exceed 3.2 billion RMB, reflecting high valuation and investor confidence in the digital economy and AI sectors.

Such capital transfers were conducted based on arm’s length negotiation between the Shareholders and did not involve any increase of registered capital by our Company.

• The company secured significant capital through Series B6-3, B6-3+, and B6-4 financing rounds, totaling over 2.45 billion RMB.
• Major institutional investors include Hangzhou Chengtou Industrial Fund, High-tech Orinno, and Zhuhai Huafa, with the latter contributing 500 million RMB.
• A significant portion of the funding was structured through convertible notes issued to entities like Lenovo Venture Capital and Tianchuang Capital.
• The convertible notes were officially converted into company shares following the firm's transition into a joint stock company in March 2025.
• The financing structure ensures that note-holders receive equity interests and rights equivalent to other Pre-IPO investors.
• The document notes a subsequent phase of equity transfers involving multiple investment groups occurring in August 2025.

The conversion of the convertible notes issued by each of Tianchuang Capital, Lenovo Venture Capital, Xiarui Investments and Zhuhai Huafa shall be effected by way of issuance of newly issued Shares at the agreed conversion prices.

• The company underwent a series of share transfers between existing shareholders and Pre-IPO investors through arm’s length negotiations.
• These capital transfers were secondary market transactions and did not involve the issuance of new registered capital by the company.
• A diverse group of investment partnerships and venture capital funds, such as Shanchuang Zhizhi and Lingtou Future, acquired significant shareholdings.
• The company's financing history progressed through eight distinct stages, from Series Angel to Series B6, spanning from 2021 to 2025.
• The company experienced exponential growth in valuation, rising from a pre-money valuation of RMB387.02 million at the Angel stage to nearly RMB20 billion by Series B6.

Such capital transfers were conducted based on arm’s length negotiation between the Shareholders and did not involve any increase of registered capital by our Company.

• The company underwent multiple funding rounds from Series Angel through Series B6, showing a rapid escalation in share price from RMB3.87 to RMB60.52.
• Valuation considerations were based on arm's length negotiations, factoring in commercialization milestones, R&D management, and expected market value.
• Pre-IPO investors received significant discounts to the offer price, ranging from 96.33% for early angel investors to 42.58% for the final Series B6 round.
• The company raised approximately RMB8,364.20 million from these investments, with 58% already utilized for R&D and capital expenditures.
• All existing shareholders are subject to a 12-month lock-up period following the Listing Date in accordance with PRC law.
• Beyond capital, the strategic benefits of these investments include industry resource networking and a market endorsement of the group's performance.

The Pre-IPO Investors’ investment demonstrated their confidence in our Group and served as an endorsement of our performance, strengths and prospects.

• Pre-IPO investments were structured through both the subscription of new registered capital and the transfer of existing shares between private parties.
• Valuation increases were driven by business development milestones, specifically the commercialization of specialist technology products and R&D progress.
• The transition from Series Angel to Series A was fueled by the generation of stable revenue from knowledge graph products and the start of large language model training.
• The launch of the GLM-10B model in September 2021 served as a primary catalyst for the valuation jump between Series A and Series B1 financing.
• Subsequent valuation growth was attributed to the release of the GLM-130B open-source model and the CodeGeeX high-performance coding model in 2022.
• Calculations for cost per share were adjusted to account for the company's conversion to a joint stock company and a planned share subdivision in 2025.

The increase in valuation from the Series A Financing to the Series B1 Financing was mainly due to the launch of GLM-10B, our first ten-billion-parameter pre-trained large model, in September 2021, which significantly raised our market visibility.

• The company's valuation surged across multiple Series B funding rounds driven by the successful launch of the ChatGLM foundation model and its open-source counterpart.
• The release of Zhipu QingYan in August 2024 marked a significant milestone as one of the first domestic large models to complete regulatory filing in China.
• Continuous R&D advancements, specifically the release of GLM-4 and GLM-4V, further solidified the company's market position and growth potential.
• Pre-IPO investors were granted extensive special rights, including liquidation priority, redemption rights, and director nomination rights.
• In preparation for the Global Offering, most special investor rights were terminated or suspended to comply with Stock Exchange listing requirements.
• The Sole Sponsor has confirmed that the Pre-IPO investments meet the regulatory standards for the upcoming listing.

The increase in valuation from the Series B4 Financing to the Series B5 Financing was mainly due to the launch of Zhipu QingYan in August 2024, which was one of the first domestic large model products to complete regulatory filing in China.

• Legend Capital and Meituan serve as the company's Pathfinder Sophisticated Independent Investors (SIIs).
• These investors held over 10% of the issued share capital for at least 12 months prior to the initial filing.
• The company maintains that these investors are independent from core connected persons and lack acting-in-concert arrangements.
• Legend Capital's involvement includes the nomination of Mr. Li Jiaqing as a Director.
• The aggregate holding of SIIs is expected to remain above 10% provided the market capitalization exceeds HK$30 billion at listing.
• Legend Capital's structure involves complex layers of limited partnerships and management companies controlled by entities like LCM and Juncheng Hezhong.

Save for being our Sophisticated Independent Shareholders and the nomination of Mr. Li Jiaqing as a Director by Legend Capital, each of our Sophisticated Independent Investors is independent from and not connected with any core connected persons.

• The Social Security Zhongguancun Innovation Fund is primarily funded by the National Social Security Fund Council, which holds a 98.04% limited partner interest.
• Legend Capital SIIs are managed by LCM, a professional investment institution with over HK$15 billion in assets under management across technology and healthcare sectors.
• Under listing guidelines, multiple shareholding entities managed by the same fund manager are aggregated as a single Sophisticated Independent Investor.
• Tianjin Sankuai, an investment arm of Meituan, maintains a diverse portfolio including major listed companies like Li Auto and Hesai Group.
• Meituan's investment portfolio saw significant growth, rising from approximately RMB33.98 billion in late 2022 to RMB41.31 billion by the end of 2024.
• Compliance with Rule 18C.05 is demonstrated through the disclosure of specific shareholding percentages at the time of the first filing and the preceding 12-month period.

The different shareholding entities are purely different funds managed by the same fund manager and should be aggregated as one SII.

• The text details the complex ownership structures of Qiming Venture Partners' funds, Qiming Rongqian and Qiming Rongkai, which are aggregated as a single Sophisticated Investor (SII).
• Qiming Venture Partners is identified as a leading Chinese venture capital firm with assets under management (AUM) growing from US$6.0 billion in 2021 to US$9.0 billion by late 2024.
• Significant state-linked involvement is noted in Qiming Rongkai, which is 30% owned by a fund controlled by the Kunshan Municipal Government State-owned Assets Supervision and Administration Office.
• Other major Pre-IPO investors holding over 1% interest include Xinglian Zhaoji and Tongzhi Investments, both ultimately controlled by Ye Xiaobin through various management entities.
• Ant Group is represented in the investment pool through its vehicles Shanghai Yunya and Shanghai Feiya, though no single shareholder holds more than one-third of Ant Group.
• The document outlines the specific shareholding percentages of these entities at key regulatory milestones, such as the date of the First Filing and the Latest Practicable Date.

Qiming Venture Partners is a leading venture capital firm in China, which is focused on investing in outstanding companies at their early and growth stages in the technology and healthcare industries.

• Shanghai Feiya is revealed to be a subsidiary of Ant Group, tracing its ownership through Cayman Islands entities and the Ant Unicorn Fund.
• The Haihe Fuxin Youda Fund is heavily backed by the Wuqing District government, with over 66% of its interests controlled by the local SASAC.
• Trend Mega and Capital Today Evergreen Fund are identified as being under the ultimate control of individual investor Xu Xin.
• Fortune Capital's investment vehicles, such as Dachen Chuanghong, are ultimately controlled by the Hunan Provincial State-owned Cultural Assets Supervision and Administration Committee.
• Huakong Capital entities, including Beijing and Qingdao branches, operate as limited partnerships ultimately controlled by Zhang Yang.
• The text details a complex web of offshore holding companies, state-owned assets, and private equity partnerships within the PRC regulatory framework.

Shanghai Feiya is wholly owned by Accelerator VIII Ltd., an exempted company incorporated in the Cayman Islands with limited liability and wholly owned by Ant Unicorn Fund, L.P.

• Zhongguancun Science City operates as a limited partnership primarily controlled by the Haidian District SASAC through a 99% limited partner stake.
• Zhuhai Huafa is a venture capital and equity investment entity ultimately governed by the Zhuhai Municipal Government SASAC.
• Zhihui Linghang involves a complex ownership structure where the Shanghai Pudong SASAC maintains ultimate control over nearly 99.8% of the partnership interests.
• The AI Fund is heavily backed by the Beijing Municipal Government, with 99% of its capital provided by a guidance fund controlled by the Beijing SASAC.
• These entities demonstrate a recurring pattern of PRC state-owned assets supervision and administration commissions (SASAC) acting as the ultimate controlling authorities of major investment vehicles.

AI Fund is owned as to (i) 0.5% by Beijing Jingguoguan Property Management Co., Ltd. as its general partner... and (iii) 99% by Beijing Municipal Government Investment Guidance Fund (Limited Partnership) as its sole limited partner.

• Tencent Investment is revealed to be a multi-layered entity ultimately controlled by Tencent founder Ma Huateng through a series of holding companies and partnerships.
• Duoxiang Network operates under a complex structure owned by Hangzhou Zhenxi, with ultimate beneficial ownership shared among several individuals including Zheng Junfang and Wu Zeming.
• Hangzhou Chengtou Industrial Fund represents a state-linked investment vehicle, ultimately controlled by the Hangzhou Municipal People’s Government.
• Beijing Shunying is a limited partnership with a diverse ownership base, including interests held by high-profile tech figure Lei Jun.
• The text details the intricate web of general and limited partnerships that define the legal and financial control of major Chinese investment entities.

The general partner of both Shenzhen Tengyuan and Shenzhen Tenglu is Shenzhen Tengqing Enterprise Management Co., Ltd., which is controlled as to 80% by Ma Huateng, the founder of Tencent.

• Prosperity7 is a protected cell company in Guernsey wholly-owned by Saudi Aramco's venture arm.
• The Daxing Industrial Fund is almost entirely controlled by the State-owned Assets Supervision and Administration Commission of Beijing's Daxing District.
• High-tech Orinno is a PRC limited partnership dominated by entities ultimately controlled by the Chengdu High-tech Zone Finance and State-Owned Assets Bureau.
• The Group reported no major acquisitions or disposals exceeding the 25% threshold defined by Listing Rules during the Track Record Period.
• A strategic consolidation occurred with Beijing Lingxin Intelligent, moving from a 5.02% stake to 100% ownership to bolster AI market share.

The acquisitions were made as part of the Group’s strategy to expand its market share of artificial intelligence in the PRC.

• The company acquired a 94.98% equity interest in Beijing Lingxin Intelligent for approximately RMB73 million, settled in October 2024.
• The acquisition was conducted through arm’s length negotiations and complied with all material PRC laws and regulatory requirements.
• Employee Ownership Platforms, Huihui and Zhideng, were established in 2021 to incentivize key personnel through partnership interests.
• Dr. Liu serves as the sole general partner for both platforms, maintaining all management and voting powers while participants hold economic interests.
• The incentive schemes include a diverse group of participants, ranging from executive directors to full-time interns serving as algorithm specialists.

The eligible participants of the Employee Ownership Platforms are only entitled to the economic interest therein.

• A formal Concert Party Agreement establishes a unified voting bloc among key founders and entities, ensuring collective control over the company.
• Dr. Liu maintains ultimate decision-making authority in the event of any internal disputes or disagreements among the controlling shareholders.
• The group of Controlling Shareholders collectively holds approximately 33.03% of the company's shares prior to the Global Offering.
• The company previously initiated a preliminary filing for an A Share Listing in mainland China, involving a tutoring agreement with CICC.
• Management pivoted to a Hong Kong listing to leverage an international platform, enhance capital access, and attract a more diverse global investor base.

In particular, Dr. Liu and Dr. Tang will exercise their shareholder rights as Shareholders in relation to major matters of the Group consistently and collectively, and in the event of a disagreement or dispute, the final decision shall rest with Dr. Liu.

• The text details the company's capitalization structure before and after a planned share subdivision, conversion of unlisted shares, and global offering.
• Controlling shareholders, including Beijing Lianpai and several individual doctors, maintain significant stakes but their shares generally do not count toward the public float.
• Sophisticated independent investors such as Legend Capital, Meituan, and Qiming Venture hold substantial positions that will convert to H shares contributing to the public float.
• A diverse group of institutional investors, including Tencent Investment and various technology funds, participate in the ownership structure with varying degrees of share conversion.
• The post-offering structure distinguishes between H shares and unlisted shares to determine the final percentage of shareholding and public market liquidity.

As of the Latest Practicable Date and immediately after completion of the Share Subdivision, the Conversion of Unlisted Shares and the Global Offering (assuming the Over-allotment Option is not exercised), the summary of the capitalization of our Company is set out as follows:

• The document details a complex cap table involving a share subdivision, conversion of unlisted shares, and a global offering.
• A diverse array of institutional investors is listed, including industrial funds like Hangzhou Chengtou and venture capital arms like Lenovo Venture Capital.
• The data tracks the transition of shareholding percentages from the 'Latest Practicable Date' to the post-offering structure.
• Most institutional holdings are designated as counting toward the 'public float,' indicating they meet regulatory requirements for liquidity.
• Public shareholders participating in the Global Offering are expected to hold 8.50% of the total shares post-transaction.
• The final total share count is projected to reach 440,230,190 shares following the completion of the offering.

Upon completion of the Share Subdivision, the Conversion of Unlisted Shares and the Global Offering.

• The company details the conversion of unlisted shares into H Shares, noting that over 10 million shares held by controlling entities will be excluded from the public float.
• Approximately 51% of total issued shares will remain as unlisted shares and will not contribute to the public float calculation upon listing.
• The expected market capitalization at listing is approximately HK$51.2 billion, with a public float representing 46.73% of total issued shares.
• The company satisfies the minimum public float requirement under Rule 19A.13A, which mandates at least 10% for this specific market valuation.
• All existing shareholders, including Pre-IPO investors, are subject to a mandatory 12-month disposal prohibition under PRC law following the listing date.
• Specific lock-up periods are defined for key persons and controlling shareholders, ranging from 6 to 12 months for various investment entities.

Pursuant to the applicable PRC law, all existing Shareholders (including our Pre-IPO Investors) are prohibited from disposing of any of the Shares held by them within the 12 months following the Listing Date.

• Key leadership figures, including co-founder Wang Shaolan and chief scientist Zhang Bo, are subject to strict share disposal restrictions under Rule 18C.14.
• The company has established Employee Ownership Platforms, Huihui and Zhideng, to grant partnership interests to executive directors and core R&D experts.
• Specific algorithm experts, including Mr. Yan Xingyu, Dr. Gu Xiaotao, and Dr. Du Zhengxiao, are identified as key personnel subject to lock-up rules.
• The company must maintain a free float market capitalization of at least HK$600 million to comply with Stock Exchange listing requirements.
• Post-offering, the company expects a free float of approximately 11.7 million H shares, valued at HK$1,363.9 million, comfortably exceeding the regulatory threshold.
• The corporate structure involves a complex network of 100% owned subsidiaries across various Chinese cities including Beijing, Shanghai, and Shenzhen.

All the above executive Directors, senior management members and key persons of our Company shall be subject to disposal restrictions pursuant to Rule 18C.14 of the Listing Rules.

• The document outlines a complex corporate hierarchy of subsidiaries across major Chinese cities including Beijing, Shenzhen, and Hangzhou, following a Global Offering.
• A group of 'Controlling Shareholders' led by Dr. Liu maintains significant influence through a concert party arrangement and various investment vehicles.
• Founded in 2019, the company has positioned itself as a leader in China's pursuit of Artificial General Intelligence (AGI) through its proprietary GLM framework.
• The company successfully transitioned from research to commercialization, ranking first among China's independent developers of general-purpose large models by 2024 revenue.
• Their technological reach is extensive, with services empowering over 8,000 institutional customers and approximately 80 million devices as of mid-2025.

We were founded in 2019 on the bold idea of pursuing innovation toward artificial general intelligence (AGI) in China.

• The development of AI is categorized into five distinct stages: pre-training, alignment/reasoning, self-learning, self-perception, and consciousness.
• The company has already achieved milestones in the first three stages, including the creation of China's first 100 billion-scale model and an AI agent with rumination capabilities.
• A Model-as-a-Service (MaaS) platform serves as the commercial foundation, offering a comprehensive portfolio across language, multimodal, and coding domains.
• The platform emphasizes flexible deployment options, including cloud, API, on-premise, and on-device hosting to ensure scalability across industry verticals.
• Safety is a core focus, with the GLM-4-9B model achieving a significantly low hallucination rate of 1.3% according to Stanford's 2025 AI Index Report.

Without human supervision, machines develop their own attitudes and emotions by observing their behavior and interpreting it themselves.

• The company utilizes a 'virtuous insight flywheel' by deploying models to understand real-world usage and refine training strategies.
• Financial growth is accelerating rapidly, with revenue increasing from RMB57.4 million in 2022 to RMB312.4 million in 2024, a CAGR exceeding 130%.
• The business model relies on an ecosystem of three pillars: open-source collaboration, robust computing infrastructure, and industry partnerships.
• Open-source engagement is a primary driver of innovation, with over 45 million model downloads and 1,000 projects created by global developers.
• Technical compatibility is a core strength, as the company's models are optimized to run across more than 40 major global chip platforms.

We see open-sourcing as a key catalyst for the vitality of our ecosystem.

• The company leverages a business partner network to drive industry-wide AI transformation through a virtuous cycle of real-world feedback and model innovation.
• AI is envisioned as a ubiquitous, infinite commodity engineered from human intellect rather than finite physical resources.
• The organization acknowledges that AI's superhuman capabilities pose profound risks to the fabric of society if not handled with extreme care.
• Safety is established as the primary priority, evidenced by the company being the only Chinese firm to sign the Frontier AI Safety Commitments at the 2024 Seoul Summit.
• A commitment to humility and transparency is emphasized, including a pledge to acknowledge mistakes and interrogate values as the technology enters uncharted territories.

Even better, the supply of this new resource could be infinite, because it is not mined from the crust of the Earth but engineered from the depth of the best human minds.

• The company is a pioneer in China's AI sector, being the first to develop large models with over 100 billion parameters and ranking first among independent Chinese developers by 2024 revenue.
• Their proprietary GLM framework distinguishes itself by integrating both unidirectional and bidirectional attention mechanisms, allowing for superior language understanding and generation.
• The firm has maintained a rapid innovation cycle, releasing advanced agents like AutoGLM-Rumination, which is capable of sustained 'thinking while working' for extended periods.
• As of July 2025, their flagship GLM-4.5 model ranked third globally and first in China across twelve industry-standard benchmark tests.
• A first-mover commercial strategy initiated in 2021 has allowed the company to build a scalable Model-as-a-Service (MaaS) platform and a diverse portfolio ranging from on-device to flagship cloud models.

In March 2025, we introduced AutoGLM-Rumination, a major breakthrough for frontier AI agents as it is able to autonomously carry out agent operations that require sustained “thinking while working” for tens of minutes.

• The GLM-4.5 model ranks as a top-tier global competitor, placing first in China and leading the global open-source market as of July 2025.
• Agentic models like GLM-Z1-Rumination and AutoGLM function as a 'brain' and 'hands' system to automate complex tasks and user interface operations.
• The multimodal portfolio includes CogView4 and CogVideoX, achieving state-of-the-art status in text-to-image and video generation with specific bilingual strengths.
• The CodeGeeX platform has seen massive developer adoption, generating an average of over 100 million lines of code daily by mid-2025.
• A Model-as-a-Service (MaaS) platform allows customers to customize agents through fine-tuning, creating a feedback loop that improves core model training.

This capability is built upon two core models: GLM-Z1-Rumination, which serves as the agent’s “brain” for deep reasoning and autonomous planning, and AutoGLM, which provides the “hands” to perceive and operate any user interface.

• The company maintains a research-centric culture with over 74% of its workforce dedicated to R&D and a portfolio of 500 papers with 58,000 citations.
• A strategic partnership with Tsinghua University's Knowledge Engineering Group (KEG) provides a robust system for technological exchange and talent recruitment.
• The 'Model-as-a-Service' (MaaS) platform offers an all-in-one ecosystem including language, multimodal, and coding models like GLM-4.5 and CodeGeeX.
• The platform is designed to democratize AI, allowing institutional customers to fine-tune models and develop agents regardless of their internal technical expertise.
• By moving away from traditional project-based AI, the company achieves economies of scale through standardized agent templates and scenario-based solutions.
• The business model focuses on 'Intelligence Empowerment,' serving over 8,000 institutional customers across sectors like finance, healthcare, and manufacturing.

This association, rooted in a shared academic lineage, helps instill continuity and coherence in our technological roadmap.

• The company pioneered the open-source movement for Chinese large models, achieving over 45 million downloads and global recognition from Stanford University.
• A robust hardware strategy ensures model compatibility with over 40 major domestic and international chips through proprietary operator libraries.
• The 'Project Z' initiative empowers early-stage startups with limited technical resources to implement large models in specialized, long-tail market scenarios.
• Model-as-a-Service (MaaS) platforms are being used to integrate AI agent capabilities into diverse sectors including finance, e-commerce, and smart vehicles.
• Strategic regional collaborations focus on building national foundation model platforms across China and Southeast Asia to drive localized AI development.

In this way, our business growth is supported not only by our own efforts, but also by the collective success of our business partners.

• The company is led by a management team with elite academic backgrounds, including experts in large language models and data mining from institutions like Tsinghua University.
• Strategic guidance is provided by esteemed advisors such as Professor Zhang Bo, an academician of the Chinese Academy of Sciences.
• The organization maintains a close collaboration with the KEG research group at Tsinghua University, focusing on original AI innovation and the GLM model framework.
• A primary business strategy involves strengthening R&D to explore the upper bounds of intelligence through iterative development of foundation models.
• Future technical goals include optimizing training infrastructure and developing self-evolution paradigms to reduce costs and boost performance.
• The company aims to create models capable of autonomous exploration and reflection to solve complex tasks in real-world environments.

We are continuously developing foundation models which can explore the upper bounds of intelligence.

• The company aims to strengthen its GLM framework by optimizing technology infrastructure and data processing platforms for scalable model deployment.
• Development is shifting toward multi-agent AI systems and embodied intelligence capable of interacting with the physical world.
• Future research goals include pushing foundation models toward understanding emotion and consciousness to create personalized superintelligent assistants.
• The Model-as-a-Service (MaaS) platform is being diversified across various parameter scales to reduce hardware barriers and democratize AI access.
• Strategic partnerships with consumer electronics and internet companies are being leveraged to integrate on-device intelligence and adaptive edge computing.

In the future, we will push the boundaries of large foundation models to understand emotion and consciousness such that we can build a personalized, superintelligent assistant for everyone.

• The company is integrating AI capabilities into its lifestyle platform to enhance translation, data analytics, and content creation.
• Project Z serves as an industry network program that provides early-stage startups with access to large models and technical infrastructure.
• A core strategic pillar involves recruiting and retaining global top-tier R&D talent through a culture of scientific inquiry.
• The business model is centered on a Model-as-a-Service (MaaS) platform that delivers intelligence across diverse computing environments.
• The MaaS platform features a comprehensive portfolio of language, multimodal, agentic, and coding models for institutional and individual use.

As a company built by scientists and engineers, talent serves as the bedrock of our sustained success and long-term growth.

• The Model-as-a-Service (MaaS) platform streamlines business workflows by offering customizable agent templates and scenario-based solutions.
• Advanced infrastructure co-design allows models to scale from 1.5 billion to 230 billion parameters across diverse hardware like mobile phones and smart vehicles.
• The platform features a user-friendly, codeless development environment that enables three-step fine-tuning without requiring in-house AI expertise.
• Significant cost efficiency is achieved through high-performance smaller models, such as GLM-4-Air, which reduces computing costs by 20 times compared to larger competitors.
• The system prioritizes speed and reliability, boasting reasoning speeds of 200 tokens per second and a low hallucination rate of 1.3%.

For example, our reasoning model GLM-4-Air, with a model size of only 32 billion parameters, delivers a performance similar to competitors’ models with a size of 671 billion parameters, reducing computing costs by 20 times.

• The Model-as-a-Service (MaaS) model allows the company to reach massive end-user bases indirectly through institutional customers.
• Growth is amplified by the commercial success of 'blockbuster' applications built by partners using the company's underlying models.
• The company focuses on general model capabilities like text generation and coding that can be applied across every industry vertical.
• A 'MaaS in the loop' flywheel creates a virtuous cycle where real-world deployment data and feedback fuel continuous model refinement.
• The technology is being integrated into consumer electronics to power on-device AI agents for real-time, context-aware interactions.
• In the technology sector, models are automating complex tasks such as document review, data analysis, and creative content generation.

The result is a virtuous insight flywheel: as our customers’ applications thrive, they produce more quality feedback that fuels further model improvements, which then allows us to optimize our models, and to attract even more institutional customers and application opportunities.

• Large-model technologies are being integrated into the public service sector to manage transportation flow, estimate arrival times, and identify structural hazards in infrastructure.
• Academic and research communities utilize AI-driven platforms for plain-language literature searches and automated summarization of scientific articles.
• Traditional corporate sectors like finance and manufacturing use these models for market trend analysis, regulatory impact assessment, and intelligent safety monitoring.
• The Model-as-a-Service (MaaS) platform provides flexible deployment through on-premise hosting for data security or cloud-based hosting for cost-effective agility.
• Revenue data shows a strong preference for on-premise deployment, consistently accounting for the vast majority of the company's total revenue across the track record period.

In road and bridge maintenance, our multimodal large models power an intelligent inspection system that collects real-time imagery and geographic data, allowing precise identification of structural hazards.

• The company offers on-premise deployment through packaged pricing based on model scale and computing resources, charged one-off or annually.
• Cloud-based services utilize both usage-based token consumption and subscription-based contracts for enterprise and individual users.
• A tiered pricing strategy aims to attract a broader customer base with lower-priced plans while securing higher margins through premium and bespoke agreements.
• Individual users access basic models for free via web and mobile apps, while advanced features are monetized through subscriptions.
• The business workflow integrates infrastructure providers to offer elastic cloud resources and localized servers for on-premise model training.
• Strategic optimization includes providing usage analytics to clients to reduce operational burdens and improve service efficiency.

We intend to introduce lower-priced plans that provide limited access to the most resource-intensive models, thereby attracting a broader customer base while encouraging clients who require premium performance to select higher-margin plans.

• The service delivery process begins with in-depth assessments of customer data requirements and application scenarios to tailor model fine-tuning.
• Deployment options are flexible, offering both cloud-based and on-premise solutions based on specific security and organizational needs.
• Project Z empowers early-stage startups with limited technical resources by providing access to large model infrastructure and technical support.
• The company's AI portfolio is structured around three core human faculties: deep thinking, cognition, and tool use.
• All specialized models, including multimodal and agent models, are built upon the foundational GLM (General Language Model) series.

In this way, our business growth is supported not only by our own efforts, but also by the collective success of our business partners.

• The GLM-4.5 flagship foundation model serves as the core for a diverse ecosystem of specialized AI, including reflection, multimodal, and coding models.
• Reflection and rumination models like GLM-Z1 are designed for 'deep thinking' to handle complex reasoning tasks beyond standard generation.
• AI agents such as AutoGLM and CoCo integrate reasoning and tool-use to perform multi-step tasks autonomously without human intervention.
• GLM-4.5 features a massive scale of 355 billion parameters, with a lightweight 'Air' version optimized at 106 billion parameters.
• As of late 2025, GLM-4.5 ranked first among global open-source models and third globally across twelve industry-standard benchmark tests.
• The model has achieved significant market traction, ranking among the top ten globally for token consumption on the OpenRouter platform.

Reflection and rumination models spend additional time “deep thinking” before generating an answer, which makes them better for complex reasoning tasks.

• GLM-4.5 achieved global popularity by ranking first on the Hugging Face trending board within 48 hours of its launch.
• The model demonstrates competitive performance across twelve industry-standard benchmarks, specifically in agentic, reasoning, and coding tasks.
• As of September 2025, GLM-4.5 recorded the lowest hallucination rate in China and the second-lowest globally on the RAG leaderboard.
• The training process involves a multilingual pre-training phase followed by three domain-specific mid-training stages to boost performance.
• Post-training alignment utilizes supervised fine-tuning (SFT) and reasoning reinforcement learning (RL) to enhance logical deduction and human interaction.
• The subsequent GLM-4.6 update further improved coding capabilities, securing a top global ranking on the CodeArena platform.

This benchmark evaluates large models based on how frequently they produce non-existent answers (i.e., hallucinations) in response to intentionally misleading questions.

• GLM-4.5 utilizes a Mixture of Experts (MoE) architecture to balance high performance with model scale efficiency.
• The model features a dual-mode design, offering a 'thinking-mode' for complex reasoning and a 'non-thinking-mode' for rapid responses.
• GLM-Z1, a reflection model, uses extended reinforcement learning to solve structured problems with high precision and self-verification.
• The GLM-Z1-Rumination model is specialized for open-ended, uncertain tasks that require iterative information processing.
• GLM-Z1 achieves a high inference speed of 200 tokens per second, significantly outperforming comparable reflection models in cost-efficiency.

This is similar to a group of physicians with different specialities working together to treat a complicated medical issue.

• GLM-Z1 is designed to run on consumer-grade hardware, lowering the entry barrier for advanced AI model usage.
• The GLM-Z1-Rumination model utilizes extended reinforcement learning to develop multi-step reasoning and logical deduction skills.
• Unlike standard reflection models, the Rumination variant integrates real-time search tools to handle open-ended research tasks and synthesize diverse information.
• CogView-4 represents a state-of-the-art text-to-image model that achieved top rankings on the DPG-Bench benchmark.
• CogView-4 is the first open-source model to support Chinese character generation and can visualize complex imagery from classical Chinese poetry.
• The multimodal suite supports ultra-long prompts and bilingual inputs, allowing for high-resolution image generation with precise semantic alignment.

This transforms the model from a static knowledge repository into a dynamic research assistant.

• The GLM-4.5V model integrates language and vision to perform complex reasoning tasks beyond simple image recognition.
• It demonstrates state-of-the-art performance in visual localization, identifying and analyzing target objects for industrial applications like quality control.
• The model can autonomously reproduce functional webpages, including static layouts and dynamic interactions, from screenshots or screen recordings.
• As a 'visual detective,' it can deduce geographic locations and coordinates by analyzing environmental clues like vegetation and architecture without external tools.
• It processes diagram-rich documents by performing OCR and text analysis simultaneously, mimicking human-like comprehension of technical reports.
• The CogVideoX series complements these visual tools by transforming creative concepts into high-quality video content.

GLM-4.5V excels as a “visual detective,” combining perceptual acuity with reasoning skills for image recognition and inference.

• CogVideoX enables high-fidelity 4K video generation at 60fps with support for complex semantic prompts and multi-channel output.
• GLM-Realtime facilitates seamless audiovisual calls with a two-minute context memory and the ability to perform vocal melodies.
• The GLM-4-Voice model utilizes an end-to-end architecture with audio tokens to eliminate information loss from text-based conversion.
• AutoGLM marks a shift from conversational AI to autonomous action by simulating human interactions with digital interfaces.
• These models integrate into hardware like smart glasses and robots to provide near-instantaneous intelligent interactions.

AutoGLM represents a major step forward in the evolution of our AI universe—from “chat” to “act,” bridging the gap between conversation-based AI and real-world task execution.

• AutoGLM 2.0 significantly outperforms leading models like GPT-4o and Claude-3.5-Sonnet in mobile and browser benchmarks.
• The system utilizes a cloud-based execution model that allows the agent to complete tasks autonomously without occupying the user's physical device.
• Development focuses on overcoming the scarcity of decision-making data by using synthesized trajectories and self-evolving reinforcement learning.
• A specialized intermediate interface disentangles high-level planning from 'grounding,' which ensures actions are based on accurate, trustworthy sources.
• The agent demonstrates high success rates (70-100%) in executing complex, multi-step tasks across different Chinese Android applications.
• AutoGLM acts as a scheduling layer that can navigate across multiple apps seamlessly, such as moving from a social platform to a reservation app.

Building capable foundation agents requires enriching them with dynamic knowledge, either through direct interaction with real-world environments or through learning from synthesized trajectories.

• GLM-PC serves as a versatile computer-use assistant capable of managing meetings, documents, and background tasks through an invisible screen feature.
• AutoGLM Rumination introduces a 'thinking while working' paradigm, combining deep reasoning with the ability to execute complex, multi-step actions.
• The Rumination model bridges the gap between research and execution, enabling the AI to perform tasks like booking travel or purchasing products after analysis.
• CoCo is an enterprise-focused agent that orchestrates corporate software, databases, and knowledge bases to automate complex business workflows.
• CoCo utilizes a personalized memory system that adapts to individual employee responsibilities to provide proactive, industry-specific insights.
• The suite emphasizes results-focused delivery, ensuring that AI interactions conclude with tangible, measurable business outcomes.

AutoGLM Rumination features “thinking while working”—it leverages outstanding reasoning capabilities powered by the GLM-Z1-Rumination model while incorporating AutoGLM’s interactive operational capabilities.

• CodeGeeX serves as a versatile AI coding assistant capable of generating 100 million lines of code daily as of mid-2025.
• The platform supports multi-language translation, automated commenting, and in-IDE technical support to maintain developer focus.
• Institutional customer growth has scaled rapidly, increasing from 48 customers in 2022 to over 3,100 by June 2025.
• Token consumption has seen exponential growth, reaching 4.6 trillion daily tokens by June 2025 compared to just 0.5 billion in late 2022.
• The GLM-4.5-Air and GLM-4.6 models maintain hallucination rates below 10% according to the HHEM-2.3 evaluation model.
• The Model-as-a-Service (MaaS) platform utilizes a lightweight approach for agent customization, offering a cost-effective alternative to from-scratch development.

As of June 30, 2025, CodeGeeX generated on average more than 100 million lines of code on a daily basis.

• A world-leading electronics producer integrated agentic models into smartphones to provide real-time audiovisual analysis and social media content generation.
• The mobile AI agent features GLM-Realtime streaming inference, allowing it to interpret on-screen text and visual details with minimal latency.
• Advanced long-context memory enables the mobile assistant to maintain coherent, multi-turn dialogues by recalling information from previous sessions.
• System-level integration via AutoGLM allows users to control device settings and initiate complex tasks through natural language commands.
• Collaboration with Kingsoft Office improved WPS AI by replacing fragmented small-scale models with a unified GLM architecture to boost scalability and accuracy.
• The transition to large-scale models for WPS AI reduced R&D costs and shortened development cycles while enhancing the quality of generated content.

This integration immediately put our AI into the pocket of a vast number of consumers, allowing them to experience and benefit from our AI at any time.

• WPS AI utilizes iterative model generation to autonomously produce outlines, slides, and speech drafts from a single theme.
• Zhaopin.com implemented an AI recruiting assistant to automate resume screening and provide post-interview evaluations for employers.
• Job seekers on Zhaopin use AI to generate and polish professional resumes based on personal strengths and preferences.
• Nieta launched a short video generation agent that transforms natural language prompts into complete videos with narration and music.
• The automotive sector is shifting from rigid command-based systems to emotionally intelligent and personalized cabin interactions.

This agent allows users to simply enter a creative idea in natural language, which is then automatically transformed into a complete short video, including visuals, narration and background music.

• A high-end electric vehicle manufacturer transitioned from rigid command execution to intuitive, natural communication using GLM series models.
• The intelligent cabin system now utilizes multi-turn guided dialogs to accurately interpret colloquial voice commands and user intent.
• Enhanced emotional intelligence allows the vehicle's AI to switch personas and provide empathetic, context-aware interactions.
• The system generates dynamic content like jokes and voice-controlled games to enrich the in-vehicle environment.
• Mengniu Dairy launched 'Mengmeng,' an AI nutritionist providing personalized health plans and proactive coaching to consumers.
• The AI Health Planner feature uses adaptive planning to dynamically adjust goals and provide empathetic motivation for users.

This allowed the system to interpret user needs with greater accuracy, significantly enhancing the comprehension of commands, even when phrased naturally or informally.

• A Chinese financial conglomerate is integrating LLMs across diverse business lines to enhance internal management and operational efficiency.
• The financial capital management LLM achieves an AI automation rate exceeding 50% for regulatory inquiries and data governance.
• Logistics-specific LLMs improve transportation planning and customer service through advanced image recognition and text extraction.
• Internal office systems are being upgraded with personal knowledge bases and 'Document + AI' workflows to boost employee productivity.
• AMiner provides AI-driven scientific intelligence for global scholars, offering academic search, literature analysis, and writing assistance.
• Zhipu QingYan serves as a general-purpose generative AI assistant capable of web search, image interpretation, and complex document reading.

Notably, this LLM achieves an AI automation rate exceeding 50% in Q&A on new financial regulations.

• Zhipu QingYan offers multimodal capabilities including video and image generation, as well as accessibility services for visually impaired users.
• The platform features a no-code environment for customizable agents, with over 800,000 active custom AI agents already created by users.
• The company maintains a core focus on R&D, driven by a team of 657 specialists in natural language processing and multimodal semantic analysis.
• Leadership includes Dr. Zhang Bo, a Chinese Academy of Sciences academician known for pioneering the theory of quotient space and granular computing.
• CEO Dr. Zhang Peng leads the practical application of the GLM model series, focusing on frameworks driven by both knowledge and data.

We are, at the core, a company of data scientists and engineers, with R&D ingrained in every aspect of what we do.

• Profiles of top technical leaders highlight expertise in large language models (LLMs), data mining, and Model-as-a-Service (MaaS) platforms.
• The leadership team includes alumni from prestigious institutions like Carnegie Mellon, UIUC, and Tsinghua University, with prior experience at Google.
• Key staff members have significant academic impact, with individual publication records exceeding 30-40 papers and thousands of citations.
• The company employs competitive remuneration and proactive retention strategies to mitigate the risk of losing specialized technical talent.
• Employment contracts strictly enforce company ownership of all inventions and intellectual property developed during the term of employment.
• Rigorous confidentiality and proprietary information clauses prevent employees from disclosing technical or trade secrets to third parties indefinitely.

We own all rights, titles and interests (including patent rights, copyrights, trade secret rights and all other intellectual property rights of any sort throughout the world) relating to any and all inventions.

• The company enforces strict non-competition clauses for staff and interns, lasting up to two years post-employment to protect intellectual property.
• Internal R&D focuses on the fundamental principles of large models, including training, parameter tuning, and algorithm optimization.
• Collaborative R&D is conducted through joint laboratories with elite universities like Tsinghua and Zhejiang to explore basic research and specific AI applications.
• The company maintains a policy of zero outsourcing for R&D activities to ensure all technological innovation remains proprietary.
• Strategic partnerships with academic leaders like Dr. Li Juanzi and Dr. Huang Minlie provide expertise in semantic mining, AI safety, and NLP.
• These academic collaborations serve as a talent pipeline and bridge the gap between foundational research and real-world system deployment.

We enter into written internship agreements with confidentiality, non-compete covenants and intellectual ownership clauses with our interns to make sure that all intellectual property developed by them in connection with their work at our Company, are our property, even after their departure from our Company.

• The organization collaborates with top-tier universities like Tsinghua and Peking University to establish laboratories focused on foundation models and ethical AI alignment.
• Strategic partnerships serve as a talent pipeline, allowing the company to identify and recruit promising AI scientists early in their careers.
• Standardized R&D agreements typically grant the company sole ownership of intellectual property and commercial rights resulting from these collaborations.
• The company provides data, materials, and funding, while academic partners provide lead researchers and meet specific project targets.
• A robust technology infrastructure is maintained to support the high computational demands of training and iterating large-scale AI models at low cost.

Through these collaborations, we are able to identify promising future AI scientists early on, ensuring a continuous supply of young talent with great potential.

• The company co-designs computing infrastructure with partners to ensure hardware is optimized for scalable model training and deployment.
• A cloud-based architecture allows for real-time processing of large data volumes and high-speed performance across diverse business operations.
• Geographically dispersed servers and rigorous contingency plans, including regular data recovery testing, ensure high availability and stability.
• The R&D process follows a disciplined, modular lifecycle from architectural design and pre-training to post-training alignment and deployment.
• Architectural strategies include investigating Mixture-of-Experts (MoE) and multimodal cross-attention designs to enhance flexibility and data processing.
• Design choices are specifically aligned with hardware parallel processing strategies to ensure training stability and reliable learning.

While we use the Transformer architecture as the foundation for most of our models, we also investigate various approaches to improve the Transformer architecture.

• The training process utilizes advanced parallelism techniques like FSDP to distribute model parameters and data across multiple computing units efficiently.
• Pre-training involves rigorous data cleaning including de-duplication, quality filtering of offensive or low-quality content, and strategic tokenization.
• Models are trained using autoregressive or span-based objectives to balance language generation with deep comprehension capabilities.
• Fine-tuning adapts general skills to specific tasks like coding or summarization using gradient-based optimization and lightweight strategies like LoRA.
• Post-training alignment is a critical final phase designed to ensure model outputs reflect human values and comply with safety standards.

First, we remove near-duplicate documents through de-duplication, which helps improve the variety of training data and reduces the risk of the model simply memorizing repeated information.

• Human-led interactions and Reinforcement Learning from Human Feedback (RLHF) improve natural conversation and instruction following.
• The Proximal Policy Optimization (PPO) algorithm is used to ensure stable behavioral adjustments and alignment with human preferences.
• Quantization techniques, including 8-bit integer formats, reduce memory usage and latency for edge device deployment.
• Continuous monitoring systems track distributional drift and error frequency to maintain post-deployment reliability.
• Controlled canary rollouts and adversarial prompt sets are utilized to ensure safety, fairness, and regulatory compliance.

PPO restricts the size of behavioral adjustments made during training by using a “clipped” objective function, helping to prevent abrupt changes and ensuring reliable progress.

• The GLM-10B model established a foundation for localized Chinese language AI and represented a breakthrough in model size and sophistication.
• The development of GLM-130B focused on lowering hardware requirements through quantization, allowing massive bilingual models to run on smaller systems.
• ChatGLM-6B democratized AI by enabling efficient performance on consumer-grade hardware for both academic and commercial use.
• The GLM-4 generation introduced multimodal capabilities and enhanced agent abilities, handling complex conversations across text and images.
• CogVideoX utilized a 3D Variational Autoencoder to compress video data to 2% of its size, significantly accelerating high-resolution video generation.
• The GLM-4-Plus model integrated Proximal Policy Optimization (PPO) to rival top international models in language understanding and instruction following.

Using a cutting-edge 3D Variational Autoencoder (3D VAE), it compresses video data to just 2% of its original size, making training faster and less resource-intensive while generating videos that are more natural and fluid.

• The release of GLM-4-Voice and AutoGLM marks a shift toward natural voice interaction and autonomous digital device control via GUIs.
• The introduction of GLM-PC and GLM-Zero-Preview demonstrates advancements in independent computer operation and complex reasoning through reinforcement learning.
• GLM-Realtime provides low-latency video comprehension and voice interaction, including the ability to perform 'Function Calls' and remember conversation context.
• AutoGLM Rumination integrates deep reasoning with practical action to handle complex research and industry-specific tasks.
• The organization outlines a three-tiered safety framework—Model, Personal, and Sovereign—to mitigate the catastrophic risks of AGI.
• Safety principles prioritize reducing hallucinations, protecting personal mental health, and collaborating with global stakeholders.

In particular, AI will seep into the finest fabric of human society and smallest crease of everyone’s lives.

• The organization has committed to the Frontier AI Safety Commitments from the 2024 AI Seoul Summit, focusing on risk management and transparency.
• A core mission is the democratization of AI, aiming to bridge the gap between the tech elite and underserved communities to ensure equitable benefits.
• Environmental sustainability is prioritized through the development of energy-efficient systems and the use of renewable energy for operations.
• The company emphasizes accountability by openly communicating challenges and implementing swift solutions for unintended consequences.
• Technical achievements include the GLM-4-9B model reaching one of the world's lowest hallucination rates at 1.3% according to Stanford's 2025 AI Index.
• The firm was the only Chinese AI company recognized in the 2024 AI Safety Index by the Future of Life Institute.

AI should be advancement for all humanity, not a privilege reserved for a few.

• The organization maintains a comprehensive management system and technical standards to protect personal information throughout its entire life cycle.
• A dedicated data security team is responsible for overseeing all protection efforts and ensuring compliance with privacy specifications.
• User data is collected only with direct authorization through self-operated applications to facilitate service delivery.
• Model training relies on a mix of open-source data for general capabilities and third-party vendor data for specialized industry knowledge.
• Third-party datasets are prioritized for high-quality, accurate, and reliable information that is often scarce in public domains.

We primarily rely on open-source data to train our models’ general capabilities, such as broad language understanding and generation, as well as basic logical reasoning.

• The company collects data directly from users via self-operated platforms only after obtaining explicit consent through clear privacy policies.
• Collected information includes basic user identifiers, device attributes, and interaction data such as AI prompts, queries, and generated content.
• Institutional client data is generally not processed unless the company is specifically commissioned as a data processor for limited service scenarios.
• Third-party data acquisition requires vendors to guarantee that information is anonymized, desensitized, and sourced from legitimate, legally compliant origins.
• Public data collection is strictly limited to non-sensitive information to ensure compliance with privacy standards and protect individual rights.

The user data we collect, store and use generally include (i) users’ basic information, such as mobile phone number; (ii) data from users’ interactions with our AI product, including user inputs such as prompts, queries, images, audio files, documents and other information, as well as the content generated by our product.

• The company processes individual user data using a minimal scope approach to ensure the least impact on data subject rights.
• Client data ownership remains with the clients, with the company acting only as an entrusted processor under strict contractual instructions.
• On-premise deployment for Model-as-a-Service (MaaS) ensures that data processing remains entirely within the client's own infrastructure.
• AI model training involves rigorous data desensitization and the removal of personally identifiable information before processing begins.
• Data sharing with third-party partners is restricted to specific, legitimate purposes and requires explicit user notification and legal undertakings.

For our MaaS clients who opt for on-premise deployment, our models are hosted within the clients’ own infrastructures, thereby ensuring that data processing remains under the clients’ control.

• The organization implements strict data storage and deletion mechanisms based on the 'minimum necessary period' principle.
• Legally mandated retention periods are strictly followed, such as keeping transaction records for at least three years.
• Retention durations are determined by assessing data sensitivity, processing risks, and specific business requirements.
• Personal information is deleted or anonymized upon account deactivation or once the processing purpose is fulfilled.
• All domestic user data is stored within China and managed through a centralized data middle platform.
• Confidential information is transmitted via API encryption to ensure data integrity and security across business units.

We store all user data collected and generated during our domestic operations in China.

• The company restricts internal data access through encryption, desensitization, and formal approval procedures on a need-to-use basis.
• Domestic PRC data remains stored within China, while overseas data from the Z.ai website is hosted in Singapore.
• Cross-border data transfers primarily involve anonymous technical data sent to domestic entities for processing, which legal advisors deem exempt from PRC export regulations.
• The Z.ai platform is designed for a global audience in English and lacks PRC-specific identifiers or geo-targeting mechanisms.
• Voluntary access by PRC users to overseas platforms is not classified as purposeful collection of personal information under current legal assessments.
• The company has not reached the thresholds requiring mandatory security assessments or standard contract filings for data export.

While PRC-based users may voluntarily access the overseas online product and submit information due to the global and open nature of online services, such possibility does not constitute purposeful collection of PRC individuals’ personal information.

• The company has established rigorous cross-border data transfer protocols, including self-assessments and security measures for transmission and storage.
• Legal advisors in both China and Singapore have confirmed that the company's data practices comply with the Personal Data Protection Act and other regional regulations.
• Third-party vendor management systems include mandatory data security assessments and contractual indemnification clauses for legal breaches.
• Operational applications undergo technical testing and privacy compliance reviews before any product launches or material updates.
• Internal security policies utilize the principle of minimum authorization, restricting employee access to sensitive user data through password complexity and bastion hosts.
• Client-entrusted data is processed only under explicit authorization, requiring clients to prove legitimate data sourcing and end-user consent.

We adopt the principle of minimum authorization for the staff who may have access to users’ personal data.

• The company utilizes a secure gateway server and comprehensive audit records to strictly manage administrative access and monitor system users.
• Data is categorized by confidentiality levels and undergoes de-identification or anonymization before being processed by personnel.
• A cybersecurity contingency plan is in place, including regular incident response drills and mandatory reporting protocols for security breaches.
• Technical safeguards include SSL/HTTPS protocols, role-based access control, firewalls, and regular data recovery testing to prevent leakage.
• The firm has achieved multiple international certifications, such as ISO/IEC 27001:2013 and DCMM Level II, validating their security infrastructure.

Data are categorized in accordance with their level of confidentiality, and de-identified or anonymized before they are processed by our personnel.

• The company maintains MLPS Level III certification, the high standard for data security and protection in the PRC.
• Internal audits confirm no material administrative penalties, investigations, or data breaches occurred during the Track Record Period.
• A comprehensive internal framework governs the entire data lifecycle, including encryption, anonymization, and contingency planning.
• A dedicated oversight committee monitors regulatory updates and maintains proactive communication with PRC authorities.
• Legal advisors confirm the group is in material compliance with all applicable cybersecurity and personal data protection laws.
• The Cybersecurity Review Measures mandate specific review triggers for critical information infrastructure operators purchasing network products.

Moreover, our service design puts great emphasis on data security, and our solutions must pass data privacy evaluations and security tests before launching or delivery to clients.

• Operators of critical information infrastructure must undergo cybersecurity reviews if their services impact national security.
• Companies with over one million users seeking foreign listings are mandated to file for cybersecurity review, though Hong Kong is currently excluded from this 'foreign' definition.
• The company maintains it is exempt from mandatory filings as it has not been officially designated as a critical information infrastructure operator by Chinese authorities.
• A specialized sales and marketing team of 145 members targets diverse industries through regional divisions in China, Hong Kong, and Singapore.
• The company utilizes an all-in-one Model-as-a-Service (MaaS) platform offering flexible deployment options including cloud, API, and on-premise solutions.
• Sales personnel collaborate with a dedicated client support team to fine-tune models for specific business environments and industry pain points.

As advised by our PRC Legal Advisors, “listing in a foreign country” should not include “listing in Hong Kong.”

• The company focuses on co-developing benchmark cases with industry leaders to demonstrate tangible service benefits and build market trust.
• Strategic collaborations with application providers integrate core AI models into existing solutions to address specific industry requirements.
• Marketing efforts rely on 'entry projects' for early adopters to cultivate a word-of-mouth reputation and establish brand equity.
• Brand awareness is driven through a mix of offline industry conferences, developer forums, and digital content marketing.
• Intellectual property is a core pillar of success, with 86 registered patents in China and over 230 additional applications in progress.
• Research and development efforts include collaborations with universities, resulting in co-owned patents for advanced LLM instruction-following.

We typically establish our presence with entry projects for specific use cases, from which we create value for the early adopters of AI model and agent solutions in each sector and cultivate our word-of-mouth reputation.

• The document lists a comprehensive portfolio of granted patents in the PRC, focusing on LLM training, web navigation, and dynamic model adjustment.
• Significant emphasis is placed on evaluation methodologies, including alignment for Chinese LLMs, safety assessments, and text quality metrics.
• The company holds 160 copyrights, with a core focus on software such as AutoGLM, Zhipu QingYan, and various iterations of the ChatGLM platform.
• Intellectual property protection is enforced through strict employment agreements featuring confidentiality, non-compete, and ownership clauses.
• The company reports a clean legal record regarding IP, with no material disputes or infringement claims during the track record period.
• The portfolio includes 314 trademarks and 59 domain names, supporting a robust Model-as-a-Service (MaaS) business infrastructure.

They acknowledge that the intellectual property developed by them in connection with their employment with us, including our in-house developed content, is our property.

• The company experienced explosive revenue growth with a CAGR exceeding 130%, rising from RMB57.4 million in 2022 to RMB312.4 million in 2024.
• Despite strong gross profit margins, the company reported massive and increasing net losses, reaching RMB2,958.0 million in 2024.
• The primary driver of these losses is aggressive investment in research and development, which surged from RMB84.4 million to over RMB2.1 billion in just two years.
• Management aims to achieve profitability by leveraging a robust portfolio of AI models and agents to capture a share of the projected US$20 trillion AI-influenced economy.
• Market forecasts suggest that by 2030, AI will empower 20% of global business decision-making and 80% of consumer smart devices.

While we achieved sustained business growth, we had loss for the year of RMB143.7 million, RMB788.0 million, RMB2,958.0 million, RMB1,235.6 million and RMB2,357.9 million in 2022, 2023, 2024 and the six months ended June 30, 2024 and 2025, respectively.

• The Chinese AI market is projected to reach RMB993.0 billion by 2030, driven by a compound annual growth rate of 35.5%.
• The Large Language Model (LLM) sub-market is expected to experience even faster growth, with a projected CAGR of 63.5% through 2030.
• The company claims a first-mover advantage as the first in China to develop self-owned models exceeding 100 billion parameters.
• R&D efforts are focused on Mixture of Experts (MoE) architecture and sparse activation to improve inference efficiency and reduce latency.
• Strategic focus is shifting toward mission-critical workflows like compliance review and decision support by reducing model hallucination rates.
• Future development plans include expanding multi-modal capabilities across text, speech, images, and video to maintain a competitive edge.

As a result, our models are expected to be reusable and verifiable in a greater number of mission-critical workflows, such as document processing, professional question-and-answer services, research and decision support and compliance review tasks.

• Development of AI agents focuses on autonomous planning, long-term memory, and reflective learning to achieve continuous self-improvement.
• Performance upgrades in code generation and content creation are designed to increase client dependency and switching costs.
• Network optimization is expected to lower unit inference costs, improving gross margins and allowing for flexible pricing strategies.
• The company's institutional customer base saw massive growth, expanding from 48 in 2022 to over 3,100 by mid-2025.
• Strategic partnerships with 'lighthouse clients' and IoT device integration are key drivers for cross-sector applicability and industry penetration.

For AI agent solutions, we are developing our agent frameworks with autonomous planning, tool use, long-term memory and reflective learning abilities, empowering the agents to deconstruct complex tasks, autonomously call external tools and self-optimize their own execution processes.

• The company is integrating AI into consumer electronics to enable cross-device collaboration and adaptive edge intelligence.
• Strategic partnerships with leading Chinese internet firms are enhancing translation, data analytics, and content creation capabilities.
• Technological innovation focuses on self-learning models and optimizing infrastructure to reduce hardware barriers for all customer sizes.
• A Model-as-a-Service (MaaS) platform allows for cost-efficient entry into diverse sectors like office software, recruitment, and dairy.
• Collaborations with market leaders provide high-quality, real-world data to refine model performance under specific hardware constraints.
• Validated specialized solutions are being standardized into a replicable framework to shorten deployment cycles across the industry.

Once validated in these specialized settings, we plan to standardize the solutions for broader use, deploying them via API, cloud and on-premise deployment.

• The company focuses on lowering implementation costs and technical risks to drive sustained enterprise engagement and recurring revenue.
• A dedicated agent workspace is being upgraded to allow customers to seamlessly integrate diverse model applications and intelligent automation.
• Short-term growth targets consumer electronics and IoT, while long-term plans involve high-barrier sectors like healthcare and education.
• Strategic partnerships allow the company to bypass the need for domain-specific expertise by collaborating with established industry leaders.
• The business model is evolving from B2B to B2C, aiming for a more diversified and resilient revenue structure through an extended intelligence network.
• R&D investment for new sectors is structured at approximately 10% to 15% of the revenue generated from those specific sectors.

This division of roles enables us to quickly understand and encode domain-specific requirements, such as pedagogical logic, assessment standards and student interaction patterns, into our solutions.

• The company employs an indirect scaling strategy by providing foundational models to clients who manage end-user operations and compliance in specialized industries.
• Technical development focuses on critical indicators for high-stakes fields, specifically targeting hallucination control, instruction following, and multi-step reasoning.
• Strategic expansion is centered on education and healthcare, where reliability and safety guardrails are paramount for sustainable revenue growth.
• The model architecture is being refined to support complex workflows like preliminary information triage and structured content generation.
• Financial data reveals a significant focus on research and development, which peaked at over 1,900% of revenue in early 2024 before beginning to stabilize.
• Achieving profitability is contingent on improving operating leverage and efficiency across R&D, administrative, and marketing expenses.

In practice, we focus on optimizing key technical indicators that are critical for these sectors, including model stability, hallucination control, instruction following and robustness in complex, multi-step task execution.

• Operating expenses surged to 869.5% of total revenue by 2024, driven by aggressive investments in R&D and marketing.
• The company is implementing cost-control measures, including simplified workflows and data-driven marketing, to improve operational efficiency.
• Research and development costs peaked as a percentage of revenue in early 2024 due to the acceleration of next-generation foundation models and multimodal agents.
• High computing service fees represent a substantial portion of R&D spending, which the company aims to mitigate through optimized model training and infrastructure.
• The Model-as-a-Service (MaaS) platform is intended to create economies of scale and shorten the time-to-market for AI applications.
• Management expects operating leverage to improve as commercialization ramps up and revenue growth begins to outpace expense increases.

With our deep academic roots as cornerstone for technological leadership, we are constantly making technological advancements and iterations of our foundation models, boosting our profitability.

• General and administrative expenses fluctuated significantly, peaking at 114.6% of total revenue in the first half of 2024.
• The company implemented share incentive schemes in late 2024 and early 2025 to attract and retain key management talent.
• Selling and marketing expenses experienced a dramatic surge, reaching 124.0% of total revenue in 2024 and 321.1% in the first half of that year.
• Management expects a temporary spike in 2025 expenses due to equity-settled share-based compensation and listing costs.
• Future strategies involve introducing digital management tools and data-driven decision support systems to refine budget and project monitoring.
• The company aims to improve operating leverage by streamlining internal finance and project management processes.

Our selling and marketing expenses as a percentage of our total revenue in the same periods, respectively, accounting for 26.4%, 81.3%, 124.0%, 321.1% and 109.3%.

• Revenue growth between 2022 and 2024 was driven by aggressive advertising investments aimed at capturing emerging market opportunities.
• Marketing expenses peaked during the launch of major AI models and Model-as-a-Service (MaaS) offerings to build brand awareness and partner ecosystems.
• The company expects marketing costs to decrease as a percentage of revenue over time through data-driven targeting and improved brand influence.
• Despite significant cash reserves of RMB2,552.0 million, the company continues to experience net cash outflows from operating activities due to R&D and net losses.
• Management plans to improve cash flow by leveraging revenue growth, optimizing operating leverage, and tightening accounts receivable collection cycles.

We strategically made more advertising investment in order to swiftly take advantage of emerging market opportunities.

• The company serves a diverse client base including enterprises, public sector entities, and individual users, experiencing rapid expansion during the track record period.
• Standard agreements cover licenses for AI models and agent solutions, hardware integration, and maintenance services with typical durations of one year.
• Intellectual property rights are strictly retained by the company, with explicit prohibitions against reverse engineering, decompiling, or disassembling their software and algorithms.
• Pricing and credit terms are determined by solution sophistication and market demand, with credit windows generally ranging from 10 to 180 days.
• The company maintains a high customer concentration, with the five largest clients accounting for 40.0% to 61.5% of total revenue across the reported periods.
• Data processing and privacy compliance are central to their service agreements, requiring client consent and adherence to applicable image and data protection laws.

No party shall conduct any reverse engineering, decompiling, disassembly or use other methods to obtain the source code and underlying algorithms of the software, hardware and related technologies provided by any other party.

• The data outlines the top customers and revenue streams for a technology firm across the fiscal years 2022, 2023, and 2024.
• Revenue sources are heavily concentrated in AI applications, telecommunications infrastructure, and smart education services.
• Credit terms for major clients vary significantly, ranging from immediate payment upon acceptance to extended periods of six months.
• The company maintains a diverse client base including publicly listed IT service firms, scientific research institutions, and government-related entities.
• Total revenue from the top five customers has shown a substantial upward trend, increasing from approximately RMB 31.8 million in 2022 to over RMB 142 million in 2024.

Customer A is an IT service company based in Beijing, China, with a registered capital of RMB868.3 million.

• The text provides a detailed breakdown of diverse customers ranging from telecommunications giants to specialized AI software developers.
• A significant portion of the client base consists of state-linked entities, higher education institutions, and government agencies in Beijing and Shenzhen.
• Financial transparency is highlighted through the disclosure of registered capital and stock exchange listings for major corporate clients.
• The company maintains a clean operational record with no material product liability claims or significant customer complaints during the track record period.
• The supply chain is categorized into four main areas: computing resources, hardware equipment, R&D support, and marketing services.

During the Track Record Period and up to the Latest Practicable Date, we did not experience any material product liability claims, product, quality issues and customer complaints.

• The company employs a rigorous internal selection process for suppliers based on quality, price, service, and credentials.
• Legal and finance departments oversee procurement protocols, primarily utilizing framework agreements for cloud computing resources.
• Standard contract terms include durations of one to four years, installment-based payments, and credit terms of up to 40 days.
• The company maintains a high level of supplier concentration, with the top five suppliers accounting for approximately 47% to 55% of total purchases.
• Computing services represent the primary procurement category, with major suppliers often having business relationships ranging from one to four years.
• Historical data shows no significant price fluctuations or material breaches of contract during the track record period.

In each year/period during 2022, 2023, 2024 and the six months ended June 30, 2025, purchases from our five largest suppliers accounted for 54.5%, 53.6%, 47.3% and 50.2% of our total purchases, respectively.

• The data details the procurement expenditures and credit terms for major suppliers across the fiscal years 2022 and 2023.
• A significant portion of the company's spending is concentrated in technical services, software development, and computing hardware.
• Supplier concentration is high, with the top five suppliers accounting for over 50% of total procurement in several periods.
• The supplier base is primarily located in Beijing, featuring a mix of private IT firms and subsidiaries of companies listed on the Shanghai Stock Exchange.
• Payment terms vary significantly between suppliers, ranging from immediate prepayments to credit windows of up to 40 days.
• Directors confirm that no major shareholders or directors hold interests in these top-tier suppliers, ensuring compliance with Listing Rules.

During the Track Record Period and up to the Latest Practicable Date, to the best knowledge of our Directors, none of our Directors, their associates or any of our current Shareholders had any interest in our five largest suppliers.

• Customer A transitioned from a non-entity in 2023 to the company's largest customer by mid-2025, while simultaneously acting as a supplier of database services.
• Revenue from Customer A surged from 1.0% in 2024 to 11.0% of total revenue in the first half of 2025.
• Supplier E, a subsidiary of a major Chinese internet firm, served as the primary supplier in 2023 and 2024 while also purchasing deployment services.
• Purchases from Supplier E peaked at RMB394.8 million in 2024, representing over 15% of the company's total procurement costs.
• The financial data reveals a complex ecosystem where major business partners occupy dual roles as both service providers and revenue sources.

Customer A, our largest customer in the six months ended June 30, 2025, was also our supplier in 2024 and 2025.

• The company maintains complex business relationships where major clients also serve as key suppliers for research, recruitment, and infrastructure.
• Customer G, a Chinese public university, contributed nearly 10% of 2024 revenue while simultaneously providing R&D services.
• Revenue from specific entities like Customer I and Customer P shows high volatility, with significant single-year spikes followed by periods of zero sales.
• Procurement from Customer Q, a cloud computing firm, has scaled dramatically from under RMB 1 million in 2022 to over RMB 94 million by mid-2025.
• Management asserts that all dual-role transactions were conducted on an arm's length basis under normal commercial terms.

Customer P/Supplier L was also one of our five largest suppliers in the same year.

• The company operates in the highly competitive Large Language Model (LLM) market, facing pressure from both independent and non-independent vendors globally.
• Key competitive factors include technical capabilities, flexible business models, ecosystem building, and the retention of high-level technical talent.
• The company emphasizes the necessity of constant innovation to mitigate risks associated with the rapidly changing AI industry landscape.
• A comprehensive list of awards from 2020 to 2024 highlights the company's recognition by major Chinese and international scientific and technological institutions.
• The company maintains all necessary material licenses and permits, including Value-added Telecommunications Business Operation Licenses, to ensure regulatory compliance.

If we are not able to upgrade, enhance or innovate our technologies and services, our business, results of operations, financial condition and prospects could be adversely affected.

• The company maintains a workforce of 883 employees, with a heavy concentration of 74.4% dedicated to research and development.
• Recruitment strategies utilize diverse channels including campus outreach and third-party recruiters to maintain high talent standards.
• Employee relations are characterized by competitive compensation, performance-based rewards, and a lack of labor union representation.
• Insurance coverage is limited to mandatory PRC requirements, notably excluding business interruption and product liability insurance.
• The company operates without owning any real property, relying entirely on leased or non-owned facilities for its business activities.

In line with general market practice, we do not maintain any business interruption insurance or product liability insurance, which are not mandatory under PRC laws.

• The company is exempted from providing a full valuation report for its land and building interests under specific Hong Kong ordinance requirements.
• Five of the company's 18 leased office properties in China lack proper title certificates, potentially allowing third parties to void the lease contracts.
• Management believes the risk of displacement is low and that relocation would be easy and inexpensive due to an abundance of available office space.
• Under the PRC Civil Code, the company can legally withhold or reduce rent if third-party claims prevent them from using their leased premises.
• Fourteen lease agreements remain unregistered with Chinese authorities, a common administrative non-compliance often caused by uncooperative landlords.
• While failure to register leases does not invalidate them, the company faces potential fines of up to RMB10,000 per property if they fail to rectify the issue.

As advised by our PRC Legal Advisors, if the relevant lessor has no right to lease the leased property and a third party other than the parties to the relevant lease contracts have legal title to such leased property, such third party may claim that the relevant lease contracts are null and void.

• The company has failed to complete the registration and filing of certain lease agreements with competent authorities.
• Management believes the lack of lease registration will not result in a material adverse effect on operations or financial standing.
• On January 16, 2025, the company and nine subsidiaries were added to the U.S. Department of Commerce's Entity List.
• Being on the Entity List restricts the company's ability to acquire items subject to Export Administration Regulations (EAR) without a license.
• Restrictions apply to U.S.-origin goods, software, and technology, as well as certain foreign-made items with U.S. content.
• The company faces potential legal and operational consequences for any violations of these U.S. export control laws.

The addition of the Listed Entities to the Entity List (the “Entity List Addition”) restricts our ability to purchase or otherwise access certain goods, software and technology (collectively, “items”) that are subject to the EAR without a license from the BIS.

• Legal counsel confirms that Entity List restrictions do not automatically extend to legally distinct subsidiaries or affiliates of listed entities.
• Non-listed entities are strictly prohibited from acting as agents, fronts, or shell companies to bypass sanctions for listed affiliates.
• Suppliers may continue providing EAR-regulated items to non-listed entities provided there is no diversion or indirect transfer to the listed parties.
• The company has shifted from purchasing physical AI chips to procuring computing resources from cloud service providers in China to maintain compliance.
• Directors believe the sanctions will not have a material adverse impact on business performance as no major investors or suppliers have withdrawn support.
• The company has established a dedicated export control compliance program to manage ongoing regulatory risks and internal controls.

However, a Non-listed Entity (or any other person) may not act as an agent, a front, or a shell company for a Listed Entity in order to facilitate transactions that would not otherwise be permissible with the Listed Entity.

• The Group has implemented a comprehensive export control compliance program based on the eight elements mandated by the BIS.
• A dedicated management committee involving legal, finance, and IT departments oversees the program to ensure organizational sustainability.
• The policy includes rigorous risk assessments to prevent 'Listed Entities' from accessing software or technology subject to EAR regulations.
• Global screening protocols are used to vet business partners against restricted party lists from the U.S., UK, EU, UN, and Australia.
• Continuous monitoring and internal audits are supplemented by external legal counsel to maintain compliance with evolving international laws.

In addition, it is our policy that EAR Items shall be prevented from being provided to the Listed Entities by Non-listed Entities.

• International Sanctions Counsel has verified that the company's export control policies provide an effective framework for mitigating risks related to the Entity List.
• The company intends to apply for removal from the U.S. Entity List once they have established a sufficient track record of compliance.
• Escalating tariffs between the U.S. and China have had limited impact because the company does not export to the U.S. or procure raw materials directly from there.
• The Group has not experienced service cancellations or material pricing adjustments despite the ongoing international trade tensions.
• The company maintains that its operations pose no material threat to the environment as it does not operate production facilities.
• ESG policies are being integrated into the business model to promote corporate social responsibility and sustainability.

We remain committed to having the Listed Entities removed from the Entity List and intend to build up the track record of our export control compliance program.

• The company's primary environmental footprint consists of office electricity usage and municipal water consumption.
• Energy and resource expenses rose from RMB0.1 million in 2022 to RMB0.8 million in 2024, reflecting business expansion and office relocation.
• Water consumption intensity increased from 4.7 to 5.7 tons per million RMB of revenue between 2024 and mid-2025 due to headcount growth.
• The firm targets a 5% reduction in electricity intensity by 2027 through stricter appliance controls and optimized R&D hardware efficiency.
• Operational strategies for sustainability include digital-first workflows, light-colored office interiors to maximize natural light, and regular manual inspections of empty rooms.
• The company reports zero Scope 1 greenhouse gas emissions, as it lacks direct production activities, with Scope 2 emissions limited to purchased electricity.

For example, we paint our offices in light colors and make sure the curtains are open during daytime to reduce use of lighting.

• The company has set a target to reduce Scope 2 greenhouse gas emissions intensity by 5% by 2027 through the adoption of energy-efficient equipment.
• Management is currently assessing the feasibility of collecting data for Scope 3 emissions to meet future Stock Exchange disclosure requirements.
• Hazardous waste production is considered negligible as the company operates out of leased offices where property managers handle disposal.
• A goal has been established to decrease non-hazardous waste intensity by 2% by 2027 compared to 2024 levels.
• The firm emphasizes a balanced lifestyle and safety training, including fire drills and mental health support, to maintain employee well-being.
• Significant investment is made in human capital, with over 1,100 hours of leadership training provided to mid-level and senior managers.

We nurtured a friendly and inspirational corporate culture that we believe is attractive to the talented scientists who are keen on our success, and we invest heavily in training and retaining them.

• The company developed an AI-powered digital sign language broadcaster for the 2022 Winter Olympics to improve accessibility for the hearing impaired.
• A systematic governance framework has been established to ensure algorithmic transparency and explainability for regulators and stakeholders.
• Model architectures are designed with clear decision paths to allow for the tracing and explanation of key AI outputs.
• Fairness and bias-mitigation strategies are integrated throughout the AI lifecycle, from data selection to continuous monitoring.
• The organization is formalizing its ESG structure by establishing a dedicated Strategy and ESG Committee and a specialized task force.

We incorporate 'explainability' as a core design principle and adopts model architectures that can provide clear decision paths so that our AI models and matching algorithms are developed in a way that allows us to trace and explain how key outputs are generated.

• The ESG task force annually evaluates environmental, social, and climate risks based on frequency, recovery costs, and social awareness.
• Specific metrics for assessment include renewable energy usage, employee travel emissions, and supplier environmental programs.
• Despite a low-impact business model primarily conducted online, the company is adopting an international-standard environmental management system.
• The company reports zero material fines or penalties related to health, safety, or environmental regulations during the track record period.
• Internal control systems are overseen by the Board of Directors, with senior management monitoring daily implementation across subsidiaries.
• Financial reporting risks are managed through standardized accounting policies for budgeting and statement preparation.

Despite that our operations do not involve any production facilities or otherwise impose any material threats to the environment, we still make our best effort to minimize our impact on the environment.

• The finance department maintains strict accounting policies through regular management account reviews and ongoing staff training.
• Legal departments conduct rigorous due diligence and contract reviews to ensure regulatory compliance and intellectual property protection.
• Internal control procedures are designed to monitor the status and timely renewal of all material licenses, permits, and government approvals.
• Human resources policies focus on high recruitment standards, performance-based remuneration, and specialized departmental training.
• A comprehensive anti-corruption and anti-bribery framework includes mandatory training for all employees and third-party agents.
• The company utilizes a dedicated whistleblowing channel to identify and mitigate internal risks related to ethical violations or illegal acts.

We maintain high standards in recruitment with strict procedures to ensure the quality of new hires and provide specialized training tailored to the needs of our employees in different departments.

• The investment department manages the full lifecycle of projects, from sourcing and due diligence to portfolio management and risk assessment.
• The company reports a clean legal record with no material administrative, arbitration, or bankruptcy proceedings during the track record period.
• A nine-member Board of Directors oversees the group, split evenly between executive, non-executive, and independent non-executive directors.
• The core leadership consists of co-founders Dr. Liu Debing, Dr. Zhang Peng, and Dr. Li Juanzi, who act in concert regarding strategic planning.
• Independent non-executive directors are appointed to provide objective advice and ensure balanced governance within the board's decision-making process.

The department sources investment projects in accordance with our investment strategy and conducts thorough pre-investment due diligence to assess the risks, business synergies and potential return of the investment projects.

• The company underwent a significant structural conversion into a joint stock company with limited liability in March 2025.
• Dr. Liu Debing, co-founder and chairman, brings 18 years of industry experience and multiple national-level science and technology awards to the board.
• CEO Dr. Zhang Peng leads business development and R&D, leveraging a 20-year career including a long tenure at Tsinghua University.
• The board includes a younger perspective through Ms. Zhang Xiaohan, an executive employee director specializing in data labeling operations.
• The governance structure was formalized in June 2025 with the re-designation of directors and the appointment of independent non-executive directors.

Ms. Zhang Xiaohan (張笑涵), aged 28, is our executive Director. She is our employee director.

• Dr. Li Juanzi, a co-founder and non-executive Director, provides strategic guidance for R&D and business development.
• Dr. Li has held a long-term academic tenure at Tsinghua University and leads several major research centers focused on knowledge intelligence and industrial IoT.
• Her career is marked by numerous high-level accolades, including the National Scientific and Technological Progress—Second Prize from the State Council.
• Mr. Li Jiaqing serves as a non-executive Director, leveraging his experience as the president of Legend Capital to guide group strategy.
• Mr. Li's professional background includes a management role at Lenovo Group and various directorships at listed logistics and wealth management firms.

Dr. Li also serves as director of the Joint Research Center between Tsinghua University’s Department of Computer Science and Technology and Siemens (China) Co., Ltd for Industrial Intelligence and Internet of Things.

• Mr. Li possesses extensive experience as a non-executive director across diverse sectors including logistics, cloud technology, and pharmaceuticals.
• The educational backgrounds of the directors are prestigious, featuring degrees from Tsinghua University, Beijing Institute of Technology, and international institutions in France and Canada.
• Mr. Wang Meng brings a strong technical and investment background, having held leadership roles at China Mobile and Ant Group.
• Dr. Yang Qiang serves as an independent non-executive director with a distinguished academic career at the University of Waterloo and Simon Fraser University.
• Dr. Yang's industry influence is significant, highlighted by his role as the founding director of Huawei's Noah’s Ark Laboratory and co-founder of 4Paradigm Inc.

He was the founding director of the Noah’s Ark Laboratory of Huawei from June 2012 to October 2014.

• Dr. Yang is a highly distinguished academic and AI expert holding multiple chair professorships and a Ph.D. from the University of Maryland.
• Dr. Yang holds the distinction of being the first Chinese fellow of the Association for the Advancement of Artificial Intelligence (AAAI).
• Dr. Xie Deren serves as an independent non-executive Director, providing high-level oversight and independent advice to the Board.
• Dr. Xie possesses extensive experience in accounting and regulatory standards, serving on committees for the Ministry of Finance and the China Securities Regulatory Commission.
• Both directors hold significant positions across various listed companies and prestigious academic institutions in China and Hong Kong.

He was the first Chinese fellow of the Association for the Advancement of Artificial Intelligence (AAAI) in 2013.

• The text details the professional background of Mr. Tang Ying, an independent non-executive Director with extensive experience in management consulting and digital marketing.
• Mr. Tang held significant leadership roles at Roland Berger and Tiens Group before founding iFORCE Beijing Interactive, which was later acquired by a Shanghai-listed company.
• The document outlines the academic credentials of Dr. Xie and Mr. Tang, highlighting degrees from prestigious institutions like Xiamen University and Shanghai Jiao Tong University.
• Standard regulatory confirmations are provided, asserting that directors have no undisclosed conflicts of interest or recent external directorships.
• The section notes a transition in corporate governance where the current Supervisor will resign upon the completion of the Listing, leaving the company without a Supervisor.

In accordance with the applicable PRC laws and regulation, the Supervisor will resign as Supervisor of the Company with effect from the completion of Listing, and our Company will no longer have any Supervisor.

• Mr. Pei Bo, a 25-year-old law graduate from Tsinghua University, serves as the Supervisor responsible for monitoring the performance of Directors and senior management.
• Dr. Zhang Peng, a co-founder and CEO, leads the Group's business development, research and development, and daily operations.
• Mr. Wang Shaolan, co-founder and deputy general manager, brings over 20 years of technology industry experience to his role in financing and operations.
• Mr. Xiao Lei, the board secretary, manages corporate governance and financing operations, drawing on a background in legal consultancy and investment banking.
• The senior management team is structured to oversee day-to-day business functions, with specific roles dedicated to legal compliance, financial strategy, and technical R&D.

Mr. Pei Bo (裴博), aged 25, was appointed as our Supervisor in March 2025. He is primarily responsible for monitoring the performance of the Directors and senior management.

• Mr. Xiao's professional background includes a vice presidency at Shanghai Soybean Network Technology and advanced legal degrees from China University of Political Science and Law.
• Mr. Cheng Ching Kit was appointed as company secretary, bringing over 12 years of experience and dual professional memberships in Hong Kong and UK governance institutes.
• The Board has established four specialized committees—Audit, Remuneration, Nomination, and ESG and Strategy—to oversee specific operational and regulatory aspects.
• The Audit Committee is chaired by Dr. Xie Deren and is composed of members with specific financial management expertise required by Listing Rules.
• Primary duties of the Audit Committee include overseeing external auditors, reviewing financial reporting procedures, and evaluating internal control systems.
• The committee is mandated to meet with external auditors without executive directors present at least biannually to ensure independent oversight.

The primary duties of the Audit Committee are (i) reviewing annually the performance of external auditors and making recommendations to the board on their appointment, reappointment, removal, fees and terms.

• The Remuneration Committee is established to formulate and supervise pay policies for directors and senior management based on performance and market benchmarks.
• Strict protocols are in place to ensure no director is involved in deciding their own remuneration to prevent conflicts of interest.
• The Nomination Committee oversees the selection criteria for leadership, focusing on cultural background, education, and professional experience.
• A key responsibility of the Nomination Committee is the annual review of the Board's structure, size, and composition to align with company strategy.
• The committees are mandated to maintain a reserve list of candidates for senior roles to ensure long-term business continuity.
• Both committees operate under specific Listing Rules and Corporate Governance Codes to ensure regulatory compliance and transparency.

The primary duties of the Remuneration Committee are (i) formulating remuneration policies for Directors and senior management based on their duties, time commitment, responsibilities and remuneration in comparable companies.

• The ESG and Strategy Committee oversees business operations, industry research, and strategic development plans.
• The company integrates ESG management into its core strategy, focusing on goals, policies, and material risk assessment.
• A formal Board Diversity Policy has been adopted to support sustainable development through a wide range of perspectives.
• Candidate selection criteria prioritize merit while considering factors like gender, age, cultural background, and professional experience.
• The current Board composition includes a diverse mix of expertise in computer science, AI research, and business management.
• The company specifically acknowledges the importance of gender diversity, currently maintaining a ratio of two female to seven male directors.

Our Company recognizes and embraces the benefits of having a diverse Board and sees increasing diversity at the Board level as an essential element in supporting the attainment of our Company’s strategic objectives and sustainable development.

• The company commits to maintaining gender diversity at all levels, explicitly stating the Board will not consist of a single gender.
• A recruitment pipeline for mid- to senior-level staff is being established to ensure future gender diversity in senior management and board succession.
• The Nomination Committee is tasked with monitoring the Board Diversity Policy and reporting progress annually in corporate governance reports.
• Remuneration for Directors and Supervisors saw a significant increase, rising from RMB 5.0 million in 2022 to RMB 43.4 million for the first half of 2025.
• The five highest-paid individuals (excluding Directors) received aggregate compensation reaching RMB 55.5 million in the first six months of 2025.
• Post-listing compensation will be guided by a Remuneration Committee based on comparable company benchmarks and individual performance.

Our Board Diversity Policy provides that our Board should not be a single gender board.

• The company has appointed Maxa Capital Limited as its Compliance Advisor to ensure adherence to Listing Rules regarding financial reporting and transactions.
• Directors and Controlling Shareholders have confirmed they hold no interests in businesses that compete with the company's operations.
• A group of eight entities and individuals, including Beijing Lianpai and several doctors, act as Controlling Shareholders via a Concert Party Agreement.
• The Controlling Shareholders' collective interest is expected to decrease from 33.03% to approximately 30.22% following the Global Offering.
• The company asserts its ability to operate independently from its Controlling Shareholders through a structured board of nine directors.
• The Compliance Advisor's term begins on the Listing Date and continues until the first full financial year's annual report is distributed.

As of the Latest Practicable Date, Beijing Lianpai, Dr. Liu, Dr. Tang, Dr. Li, Dr. Xu, Dr. Zhang, Huihui and Zhideng, by virtue of the Amended Concert Party Agreement entered into among them, were collectively interested in approximately 33.03% of the Shares.

• The Group's leadership is anchored by co-founders Dr. Liu and Dr. Zhang, who manage strategic planning, R&D, and daily operations.
• Directors are bound by fiduciary duties that mandate prioritizing the company's interests and abstaining from voting on transactions involving personal conflicts.
• The Board includes three independent non-executive directors to ensure impartial decision-making and oversight of critical corporate matters.
• The Group maintains full operational independence from its Controlling Shareholders, possessing its own licenses, facilities, and intellectual property.
• Administrative functions such as finance, HR, and legal are managed by a dedicated internal team separate from the Controlling Shareholders.
• A diversified base of customers and suppliers ensures the business can function without reliance on the Controlling Shareholders' networks.

In the event that there is an actual or potential conflict of interest arising out of any transaction to be entered into between our Group and any of the Directors or their respective close associates, the interested Director(s) shall abstain from voting.

• The company maintains full operational independence by managing its own administrative, compliance, and secretarial functions without reliance on controlling shareholders.
• All full-time employees are recruited independently through internal referrals, campus recruitment, and third-party sources rather than through parent entities.
• Financial independence is established through separate accounting systems, independent bank accounts, and a dedicated treasury department.
• The group demonstrates its ability to secure third-party funding, evidenced by previous independent Pre-IPO investments and a lack of outstanding loans from controlling shareholders.
• New corporate governance measures include amended Articles of Association that mandate directors to abstain from voting on transactions where they have a conflict of interest.
• The Board of Directors will maintain a balanced composition where at least one-third are independent non-executive directors to ensure impartial decision-making.

In particular, our Articles of Association provide that, in reviewing a connected transaction or related party transaction, the connected or related Director shall make appropriate disclosure to the Borad and shall not vote on the resolution.

• The company has appointed Maxa Capital Limited as a Compliance Advisor to ensure adherence to Listing Rules and corporate governance standards.
• Internal control mechanisms have been established to identify and manage connected transactions with controlling shareholders.
• Independent non-executive Directors are mandated to conduct annual reviews of continuing connected transactions to ensure they remain fair and reasonable.
• The Directors assert that existing corporate governance measures are sufficient to protect minority shareholders and manage potential conflicts of interest.
• The document details the substantial shareholding interests of key individuals, including Dr. Liu and Dr. Tang, both before and after the Global Offering.
• A significant portion of the shareholding is held through joint interests, reflecting a complex structure of control among the primary stakeholders.

Based on the above, our Directors believe that there are sufficient and adequate corporate governance measures in place to manage any existing and potential conflicts of interest.

• The document details the ownership structure of the company both before and after the completion of a Share Subdivision and Global Offering.
• Key institutional investors include major tech-related entities such as Meituan, Ant Group, and Tsinghua Asset Management.
• Mr. Chen Hao and associated entities like Junqi Jiarui and Junlian Xiangdao maintain a significant interest, holding over 12% of H Shares post-listing.
• A Concert Party Agreement exists between several individuals and entities, including Dr. Liu and Dr. Tang, requiring them to act in unison when exercising shareholder rights.
• The transition from Unlisted Shares to H Shares is a primary feature of the post-listing capital structure for many substantial shareholders.
• All reported interests are classified as long positions, indicating a commitment to the company's future value.

They will act in concert when exercising their shareholder rights as Shareholders of the Company.

• The document details complex ownership chains involving Legend Capital Management (LCM) and its various investment vehicles like Junlian Xiangdao and Junlian Jinfan.
• Major tech conglomerates including Meituan and Ant Group are identified as significant indirect shareholders through wholly owned subsidiaries.
• Academic and institutional involvement is highlighted by Tsinghua University's 100% ownership of Tsinghua Technology through its asset management arm.
• The disclosure establishes 'deemed interests' under the Securities and Futures Ordinance (SFO) for individuals like Mr. Chen Hao and Ms. Xu Xin.
• The text confirms that no other persons will hold a 10% or more interest in the company's voting share capital following the Global Offering.

Therefore, under the SFO, each of Mr. Chen Hao, Junqi Jiarui, Juncheng Hezhong, LCM and Lhasa Junqi is deemed to be interested in the Shares held by Junlian Xiangdao, Junlian Jinfan and Social Security Zhongguancun Innovation Fund.

• The company plans a one-for-ten share split prior to the Global Offering, reducing the nominal value per share from RMB1.00 to RMB0.10.
• Post-offering, the share capital will consist of Unlisted Shares and H Shares, with Unlisted Shares maintaining a majority stake of approximately 51%.
• Unlisted Shares and H Shares are considered the same class of ordinary shares and rank pari passu, particularly regarding dividends and distributions.
• H Shares are generally restricted from being traded by PRC persons, except for specific qualified institutional and domestic investors.
• The conversion of Unlisted Shares into H Shares for overseas trading requires strict adherence to internal approvals and CSRC filing procedures.
• Dividends for H Shares will be declared in Renminbi but can be paid in either Hong Kong dollars or Renminbi.

Unlisted Shares and H Shares shall rank pari passu with each other in all respects and, in particular, will rank equally for dividends or distributions declared, paid or made.

• The conversion of unlisted shares into H shares requires compliance with both overseas exchange regulations and PRC regulatory filings, specifically with the CSRC.
• A total of 178,282,205 unlisted shares held by 57 shareholders are slated for conversion, representing over 40% of the company's total issued shares post-offering.
• The 'Full Circulation' program allows domestic unlisted shares to be converted and traded on the Stock Exchange, either during an IPO or as a separate application.
• Post-conversion procedures involve instructing the H Share Registrar and ensuring the shares are eligible for deposit and settlement via HKSCC and CCASS.
• Under PRC Company Law, shares issued prior to a public offering are subject to a statutory one-year lock-up period from the date of listing.

The PRC Company Law provides that in relation to the public offering of a company, the shares issued prior to the public offering shall not be transferred within a period of one year from the date on which the publicly offered shares are listed on any stock exchange.

• The document outlines regulatory compliance for H-share companies regarding the 'Full Circulation' program for domestic unlisted shares.
• Directors have been granted general mandates to issue, sell, or transfer treasury shares following the completion of the Global Offering.
• Financial reporting follows IFRS Accounting Standards, which the company notes may differ materially from other jurisdictions like the United States.
• The company identifies as a leading Chinese AI developer focused on Artificial General Intelligence (AGI) and general-purpose large models.
• A proprietary Model-as-a-Service (MaaS) platform is used to deliver scalable AI solutions to over 8,000 institutional customers.
• Revenue growth is driven by flexible deployment options, including cloud-based and on-premise solutions for diverse enterprise needs.

We were founded in 2019 on the bold idea of pursuing innovation toward artificial general intelligence (AGI) in China.

• The company experienced exponential revenue growth, rising from RMB57.4 million in 2022 to RMB312.4 million in 2024.
• Research and development expenses have surged dramatically, reaching RMB2,195.4 million in 2024 as the firm prioritizes in-house AI technology.
• A significant portion of R&D spending is dedicated to computing service fees, which exceeded RMB1.5 billion in 2024 alone.
• The company plans to allocate approximately 70% of its Global Offering proceeds toward strengthening general-purpose large AI model capabilities.
• Operational efficiency is heavily dependent on the unit price of computing power, which is forecasted by Frost & Sullivan to decline in the future.
• Revenue growth is driven by the expansion of Model-as-a-Service (MaaS) offerings, including the release of state-of-the-art models like GLM-4-Plus.

Technological leadership requires sustained substantial investment in R&D.

• The company has introduced specialized 'reflection' and 'rumination' models (GLM-Z1) designed for deep thinking and complex reasoning tasks.
• A diverse portfolio of parameter scales is being developed to ensure compatibility across various hardware and computing power levels.
• The sales strategy focuses on penetrating regional markets and partnering with industry leaders to create benchmark case studies.
• Marketing expenses have scaled dramatically from RMB15.1 million in 2022 to RMB387.5 million in 2024 to support rapid growth.
• The business model relies on integrating core AI capabilities into the solutions of established industry application providers.

Reflection model and rumination model both spend additional time “deep thinking” before generating an answer, which makes them better for complex reasoning tasks.

• The AI market in China is experiencing explosive growth, rising from RMB52.4 billion in 2020 to a projected RMB993.0 billion by 2030.
• Global AI integration is expected to empower 20% of business decision-making and 80% of consumer smart devices by 2030, creating over US$20 trillion in value.
• The company identifies as the largest independent LLM vendor in China by revenue as of 2024, leveraging R&D and talent as core advantages.
• Operational success is heavily contingent on macroeconomic conditions, global demand for general-purpose large models, and evolving regulatory environments.
• Financial reporting follows IFRS Accounting Standards, reflecting the company's transition from a limited liability company to a joint stock limited liability company in 2025.

According to Frost & Sullivan, by 2030, AI will empower at least 20% of daily business decision-making worldwide and enable at least 80% of consumer mainstream smart devices globally, achieving over US$20.0 trillion in value.

• Management must use historical experience and various factors to make judgments and estimates that affect reported financial amounts.
• Revenue is recognized when control of a product or service is transferred to the customer, excluding taxes collected for third parties.
• The company distinguishes between acting as a principal or an agent based on whether it controls the service before delivery.
• On-premise deployment revenue is recognized upon delivery and customer acceptance at a specific point in time.
• Cloud-based deployment revenue is recognized over the contract term, either ratably for subscriptions or based on utilization for usage-based contracts.

Control refers to our ability to direct the use of and obtain substantially all of the remaining benefits from the product or service.

• Interest income is calculated using the effective interest method, applied to gross carrying amounts unless an asset becomes credit-impaired.
• Government grants are treated as deferred income and recognized as expense reductions in the periods the related costs are incurred.
• Financial instruments with redemption obligations are recorded as liabilities at the present value of the highest possible settlement outcome.
• Convertible bonds are designated as financial liabilities measured at fair value through profit or loss (FVPL).
• Expected credit losses (ECLs) are calculated as probability-weighted estimates of cash shortfalls for assets measured at amortized cost.
• Loss allowances generally equal lifetime ECLs, though low-risk instruments may be measured based on a 12-month default window.

A contract that contains an obligation for the Company or the Group to purchase its own equity instruments for cash or another financial asset gives rise to a financial liability even if the Company’s or the Group’s obligation to purchase is conditional on the counterparty exercising its right to redeem.

• Loss allowances for trade receivables and contract assets are consistently measured based on lifetime Expected Credit Losses (ECLs).
• Significant increases in credit risk are determined using a mix of historical experience, qualitative analysis, and forward-looking information.
• Financial assets are assumed to have increased in risk automatically if they become past due.
• A financial asset is classified as credit-impaired when specific detrimental events occur, such as debtor bankruptcy or contract breaches.
• Assets are written off entirely when there is no realistic prospect of recovery due to a lack of debtor assets or income sources.

The gross carrying amount of a financial asset or contract asset is written off to the extent that there is no realistic prospect of recovery.

• The company evaluates non-current assets for impairment by comparing their carrying amounts to their recoverable amounts, defined as the higher of fair value or value in use.
• Determining value in use involves significant management judgment regarding future cash flows, revenue levels, and operating costs discounted to present value.
• Impairment testing is conducted at the level of cash-generating units (CGUs) for property, equipment, and intangible assets, with goodwill impairments being irreversible.
• No impairment losses were recognized during the Track Record Period as recoverable amounts consistently exceeded carrying amounts.
• Financial data reveals a massive surge in research and development expenses, which reached over 700% of total revenue by 2024.
• Despite growing revenue, the company faces substantial operational losses driven by aggressive spending in R&D and marketing.

In determining the value in use, expected future cash flows generated by the asset are discounted to their present value, which requires significant judgment relating to the level of revenue and amount of operating costs.

• The company reports significant and increasing net losses, reaching over 2.3 billion RMB for the first half of 2025.
• A major driver of reported losses is the change in carrying amounts of financial instruments issued to investors, which are non-cash in nature.
• Management utilizes a non-IFRS 'adjusted loss' measure to provide a clearer view of operating performance by excluding specific non-cash and one-time items.
• Adjustments to the net loss include equity-settled share-based compensation, listing expenses, and valuation changes of investor-held instruments.
• Redemption rights issued to Pre-IPO investors are expected to terminate and convert into equity upon the completion of the Global Offering.

These redemption rights issued will be terminated and converted into equity upon the Global Offering.

• The company generates revenue through its Model-as-a-Service (MaaS) platform, offering a matrix of large models and agentic tools.
• On-premise deployment is the dominant revenue stream, accounting for over 80% of total revenue as of 2025, driven by demand for data security and regulatory compliance.
• The Internet and technology sector remains the largest industry vertical for on-premise services, though public services and telecommunications show significant growth.
• Cloud-based deployment is offered on a subscription or usage basis, but the company cannot track user industries due to privacy-focused registration processes.
• While historically focused on mainland China, the company began expanding its on-premise services to overseas markets, specifically Southeast Asia, in 2024.

When registering for our cloud-based services, users are not required to disclose their respective industries, and we do not proactively collect or retain such information in the course of service usage.

• The company utilizes cloud-based deployment to offer customers agile access to foundation models and AI tools with minimal up-front investment.
• Cost of revenue is primarily driven by payroll for service personnel and computing service fees paid to third-party cloud providers for model training and inference.
• Financial data shows a significant shift in cost structure, with computing service fees rising from 0% in 2022 to over 37% of total costs by mid-2025.
• The business maintains a presence in Southeast Asia, Hong Kong, and the United States, though regional revenue concentrations vary.
• On-premise deployment remains a major cost segment, but cloud-based deployment costs have grown rapidly as the company scales its AI offerings.
• Technical service fees are managed by outsourcing labor-intensive tasks like data annotation to third-party providers to maintain cost efficiency.

Each such invocation consumes computing power to process and analyze the relevant inputs and generate the corresponding outputs, resulting in computing service fees primarily representing amounts paid to third-party cloud service providers.

• On-premise deployment remains the primary driver of gross profit, benefiting from high-value contracts with large enterprises and public sector clients who prioritize data security.
• Cloud-based deployment margins have experienced a significant decline, dropping from a high of 76.1% in 2022 to a slight loss in the first half of 2025.
• Research and development expenses have surged dramatically, growing from approximately 84 million RMB in 2022 to over 2.1 billion RMB in 2024.
• Computing service fees have become the dominant R&D cost component, accounting for over 70% of the total R&D budget by 2025 as the company scales its AI model training.
• Payroll costs and technical service fees represent secondary R&D expenditures, reflecting the ongoing investment in specialized personnel and outsourced data cleansing.

These customers are often willing and able to pay higher prices for dedicated deployment, dedicated model services and comprehensive on-site support.

• The company is aggressively investing in foundation models and advanced training infrastructure to capitalize on the AI industry's expansion.
• Frequent, large-scale model training and retraining require significant computing power sourced from third-party suppliers.
• Research and development costs are currently recognized as immediate expenses rather than capitalized assets, as specific feasibility and benefit criteria have not yet been met.
• Selling and marketing expenses have surged dramatically, driven largely by a massive increase in advertising and marketing spend from 2022 to 2024.
• Payroll costs for sales personnel and technical service fees represent significant portions of the operational budget alongside computing service fees.
• The financial data reveals a shift from payroll-dominated marketing costs to a strategy focused heavily on broad advertising and market reach.

We are constantly experimenting with larger models, richer datasets and more frequent iterations which need substantially more computing power from third party suppliers.

• General and administrative expenses are primarily driven by payroll costs, which surged from RMB20.9 million in 2022 to RMB124.7 million in the first half of 2025.
• Payroll costs consistently represent the largest share of administrative spending, reaching 67.3% of the total administrative budget by mid-2025.
• Finance costs include significant transaction costs related to financial instruments issued to investors, which peaked at RMB31.9 million in the first half of 2025.
• The company experienced a notable shift in associate performance, moving from a loss of RMB0.5 million in 2023 to a profit of RMB21.3 million in 2024.
• Fair value changes of financial instruments measured at FVPL showed substantial growth, peaking at RMB66.3 million in 2024.

We recognized impairment loss on trade and other receivables and contract assets of RMB0.03 million, RMB19.8 million, RMB17.0 million, RMB0.8 million and RMB10.9 million in 2022, 2023 and 2024 and for the six months ended June 30, 2024 and 2025, respectively.

• The company classifies Pre-IPO financial instruments as liabilities due to redemption rights granted to investors, which will terminate upon certain future events.
• Significant losses were recorded in the carrying amount of these financial instruments, reaching RMB429.3 million for the first half of 2025.
• Total revenue experienced a massive 325.0% increase year-over-year, driven primarily by the expansion of large model services.
• On-premise deployment revenue surged by 503.5%, reflecting a strategic shift toward providing complex services to a broader customer base.
• Cost of revenue rose by 315.9%, largely due to increased computing service fees and payroll costs associated with business scaling.
• Gross profit margin improved slightly to 50.0%, with on-premise deployment maintaining a robust margin of 59.1%.

Our revenue derived from on-premise deployment increased by 503.5%, from RMB26.8 million for the six months ended June 30, 2024 to RMB161.8 million for the six months ended June 30, 2025 as we continued to provide more complex large model services to our customers.

• Cloud-based deployment shifted from a gross profit of RMB5.2 million to a gross loss of RMB0.1 million as the company cut prices to capture market share.
• Research and development expenses surged by 85.6% to nearly RMB1.6 billion, driven by massive investments in computing power for foundation model iteration.
• Operating expenses across selling, marketing, and administration rose significantly, primarily due to increased payroll costs associated with rapid business expansion.
• Finance costs and carrying amounts of financial instruments increased sharply following new equity financing rounds and transaction costs.
• The company's total loss for the six-month period nearly doubled, growing from RMB1.2 billion in 2024 to over RMB2.3 billion in 2025.

Our research and development expenses increased by 85.6%, from RMB859.2 million for the six months ended June 30, 2024 to RMB1,594.7 million for the six months ended June 30, 2025.

• Total revenue grew by 150.9% to RMB312.4 million in 2024, driven by model iterations and expansion into broader industry sectors.
• Cloud-based deployment revenue surged by 306.6%, though its gross profit margin plummeted from 31.0% to 3.4% due to strategic price reductions.
• Research and development expenses saw a massive 315.1% increase, reaching over RMB2.1 billion as the company invested heavily in foundation models.
• Computing service fees for R&D alone increased by RMB1.24 billion, reflecting the high cost of third-party computing power for model training.
• Overall gross profit margin declined from 64.6% to 56.3%, primarily impacted by the rising costs of service personnel and cloud infrastructure.

Our gross profit margin for cloud-based deployment decreased from 31.0% to 3.4%, as we continued to lower service prices strategically and in line with the market trend.

• Selling and marketing expenses surged by 282.9% to RMB387.5 million, driven by aggressive advertising and a significant expansion of the sales team.
• General and administrative costs doubled due to increased hiring and share-based payment expenses as the company scaled its operations.
• The company's annual loss widened dramatically by 275.4%, reaching nearly RMB3 billion in 2024 despite growth in specific financial instruments.
• Revenue growth was robust at 116.9% between 2022 and 2023, fueled by the deployment of increasingly sophisticated large AI models.
• Cloud-based deployment revenue saw a massive 359.7% increase, reflecting a strategic shift toward more versatile and accessible service models.

As a result of the above factors, our loss for the year increased by 275.4%, from RMB788.0 million in 2023 to RMB2,958.0 million in 2024.

• Gross profit surged by 156.6% to RMB80.5 million in 2023, driven by a significant margin increase in on-premise deployment services.
• Cloud-based deployment margins collapsed from 76.1% to 31.0% due to intense market competition and strategic price reductions aimed at capturing market share.
• Research and development expenses skyrocketed by 526.8%, reaching RMB528.9 million as the company invested heavily in foundation model iteration and computing power.
• Selling and marketing costs increased nearly sixfold to RMB101.2 million to capitalize on emerging market opportunities through aggressive advertising.
• General administrative expenses and impairment losses also rose, with the latter seeing a sharp spike due to a one-time write-off of other receivables.

We had also made strategic reductions to our service prices to expand our market share and attract additional customers.

• Finance costs surged from RMB5.7 million to RMB26.3 million between 2022 and 2023, driven by increased lease liabilities and financial advisor fees.
• The company experienced a dramatic 448.5% increase in annual losses, reaching RMB788.0 million in 2023 compared to RMB143.7 million the previous year.
• Fair value changes in financial instruments and carrying amounts of instruments issued to investors saw massive spikes, largely due to additional equity financing activities.
• The consolidated statement of financial position reveals a significant net liability position that widened to over RMB6.1 billion by mid-2025.
• Current liabilities are heavily dominated by financial instruments issued to investors, which grew from RMB457.9 million in 2022 to over RMB9.5 billion by June 2025.

As a result of the above factors, our loss for the year increased by 448.5%, from RMB143.7 million in 2022 to RMB788.0 million in 2023.

• The company experienced a massive increase in net current liabilities, rising from RMB232.3 million in 2022 to over RMB8.2 billion by late 2025.
• The primary driver of these liabilities is the issuance of financial instruments to investors, which reached RMB9.9 billion by October 2025.
• Management is addressing liquidity concerns by securing banking facilities and implementing disciplined working-capital management to optimize debt structures.
• Property and equipment assets saw a dramatic 1,801.4% surge in 2023, largely due to the procurement of computing hardware and electronic equipment.
• Despite significant cash reserves of over RMB2.5 billion, the company maintains a negative net current asset position due to the scale of its short-term obligations.

Our net current liabilities increased by 651.3% from RMB232.3 million as of December 31, 2022 to RMB1,745.4 million as of December 31, 2023, primarily attributed to an increase of RMB2,721.9 million in financial instruments issued to investors.

• The company experienced a significant shift in its operational model, moving from owning to leasing electronic equipment and computing hardware to improve efficiency.
• Property and equipment values fluctuated due to business expansion, right-of-use asset increases, and subsequent depreciation cycles.
• Intangible assets saw a 159.5% surge in 2023, driven primarily by the acquisition of a subsidiary and the integration of its patent rights.
• Goodwill remained stable at RMB39.4 million following the 2023 acquisition of Beijing Lingxin Intelligent, with no impairment losses recorded through mid-2025.
• The company utilizes an extended eight-year forecast period for goodwill impairment testing, reflecting the long-term growth trajectory of its underlying technology.

This is because we partially shifted from owning to leasing electronic equipment for operational efficiency.

• The company conducted goodwill impairment tests for Beijing Lingxin Intelligent, concluding that no impairment was necessary through mid-2025.
• Key financial assumptions include annual revenue growth rates up to 35% and steady gross profit margins between 54% and 58%.
• Sensitivity analysis indicates that a 2% to 3% increase in the pre-tax discount rate would eliminate the existing financial headroom.
• Interests in associates, specifically the Xinglian equity investment fund, saw a significant increase from nil in 2022 to RMB290.3 million by June 2025.
• Other non-current assets grew substantially, driven by input VAT deductions and prepayments for computing service fees.

The following table sets out the hypothetical change to pre-tax discount rate that would have removed the remaining headroom.

• Non-current assets more than doubled from RMB97.3 million to RMB202.6 million in the first half of 2025.
• The surge in assets was primarily driven by a significant increase in input VAT to be deducted, resulting from expanded business procurements.
• Prepayments for computing service fees rose sharply as the company scaled its procurement of computing power.
• Contract fulfillment costs, representing expenses for ongoing service contracts, reached RMB70 million by mid-2025.
• A portion of service fees for on-premise solutions is being withheld by customers until warranty periods or specific conditions are met.
• Inventory write-downs have steadily increased each year, reaching RMB2.18 million by June 30, 2025.

We had prepayment of computing service fees in 2024 as we substantially increase our procurement of computing power.

• The company measures impairment by comparing expected consideration against direct costs not yet recognized as expenses under IFRS 15.
• Inventories are valued at the lower of cost or net realizable value, with hardware costs determined by transaction prices.
• Impairment assessments for signed contracts rely on binding agreements and historical execution data for cost projections.
• Write-downs for inventories and contract costs have steadily increased from RMB0.6 million in 2022 to RMB2.2 million by mid-2025.
• Total inventory and contract costs surged by 109.0% in the first half of 2025, reaching RMB67.9 million due to expanded on-premise deployment services.
• Average turnover days for these assets fluctuated significantly, peaking at 162 days in 2023 before dropping to 95 days by June 2025.

Our inventories and contract costs increased by 109.0% from RMB32.5 million as of December 31, 2024 to RMB67.9 million as of June 30, 2025, primarily due to an increase in contract fulfillment cost of RMB37.1 million primarily due to the completion and acceptance of certain ongoing on-premise services.

• Inventory turnover days fluctuated significantly, peaking at 162 days in 2023 due to service complexity before dropping to 82 days in 2024 as cloud-based deployments increased efficiency.
• Total trade and other receivables experienced an explosive 1,393.4% increase between 2022 and 2023, driven by capital contributions and hardware procurement facilitation.
• The company increasingly acts as a procurement agent for customers, advancing payments to suppliers for hardware which significantly inflates 'other receivables' on the balance sheet.
• Trade receivables grew from RMB6.6 million in 2022 to RMB171.7 million by mid-2025, reflecting rapid business scaling and a corresponding rise in credit loss allowances.
• A substantial portion of the 2023 receivables spike was attributed to RMB120.3 million in capital contributions from Pre-IPO Investors.

Our average inventory and contract costs turnover days increased from 124 days in 2022 to 162 days in 2023, primarily due to the complexity and diversification of our services.

• The company experienced a significant increase in trade receivables from RMB6.6 million in 2022 to RMB145.8 million by mid-2025, driven by business expansion and a broader service portfolio.
• A major settlement of RMB192.8 million in other receivables occurred by October 2025, resolving potential recoverability concerns regarding hardware-related transactions.
• The average trade receivables turnover days increased sharply from 21 days in 2023 to 112 days in the first half of 2025, reflecting shifts in customer payment practices and milestone-based billing.
• Loss allowances for trade receivables have been scaled up significantly, reaching RMB25.9 million by June 30, 2025, to account for expected credit losses under IFRS 9.
• On-premise deployment services utilize a milestone-based payment structure, which contributes to the varying aging profile of the company's receivables.

Our average trade receivables turnover days remained relatively stable at 23 days in 2022 and 21 days in 2023.

• Trade receivables turnover days increased significantly from 58 days in 2024 to 112 days by mid-2025.
• Management maintains confidence in debt recoverability due to the strong credit history of private and state-owned enterprise clients.
• Short-term investments (FVPL) surged to RMB549.4 million in 2025, driven largely by the purchase of structured deposits.
• The company utilizes principal-guaranteed wealth management products to improve cash utilization without disrupting operations.
• Strict internal controls and risk management policies govern investment decisions, including board approval and issuer diversification.

Consequently, our trade receivables turnover days increased notably to 58 days in 2024 and further increased to 112 days for the six months ended June 30, 2025.

• The company maintains a strategy of investing surplus cash into low-risk, short-maturity wealth management products to maximize capital efficiency.
• Total trade and other payables saw a massive surge from RMB 25.8 million in 2022 to RMB 838.4 million by mid-2025.
• Computing service fees represent the largest and fastest-growing liability, reflecting intensive R&D and operational scaling.
• The average trade payables turnover days increased significantly from 7 days in 2022 to 120 days in 2025, indicating a shift in payment cycles.
• A temporary spike in 2023 payables was partially attributed to RMB 97.8 million in advances from Pre-IPO investors prior to capital injection.

In the future, we may continue to purchase low-risk wealth management products with a short maturity period based on surplus cash situation to maximize our capital utilization efficiency.

• Trade payables turnover days increased significantly from 7 days in 2022 to 120 days by mid-2025, reflecting improved bargaining power and supply chain management.
• As of October 31, 2025, the company had successfully settled 63.6% of its outstanding trade payables from the June 30 balance.
• Contract liabilities, representing customer advance payments, grew by 110.2% in 2023 due to business expansion before stabilizing at approximately RMB75 million.
• Deferred income from government grants for AI research and development increased steadily, reaching RMB36.2 million by June 30, 2025.
• Government grants are recognized as a reduction of expenses in profit or loss, with a peak of RMB36.3 million charged to profit or loss in 2024.
• The timing of revenue recognition for sales contracts caused fluctuations in contract liability balances throughout the Track Record Period.

Our average trade payables turnover days increased to 120 days for the six months ended June 30, 2025 primarily due to enhanced supply chain management and bargaining power.

• The company confirms sufficient working capital for at least 12 months based on current cash, investments, and expected IPO proceeds.
• Cash burn rate is defined by operating activities, equipment purchases—primarily computing power—and lease payments.
• Historical cash burn has escalated dramatically from RMB3.0 million in 2022 to RMB327.3 million for the first half of 2025.
• Total liquidity as of October 2025 stands at RMB8,943.1 million, with an additional RMB3,785.3 million expected from the Global Offering.
• Financial viability is estimated to last between 27.3 and 38.9 months depending on the inclusion of various IPO proceeds.
• Future cash management strategies include stricter client credit assessments and reinforced receivable collection efforts.

To be specific, a major portion of our expenditure on equipment purchases and leasing is related to computing power equipment, which is essential in our day-to-day operations and research and development activities and representing significant cash outflows.

• The company implements strict creditworthiness procedures and remedial measures to manage accounts receivable and ensure timely client payments.
• Operating activities consistently resulted in net cash outflows throughout the track record period, reaching a peak deficit of RMB2,244.9 million in 2024.
• Net cash used in operations is primarily driven by significant losses before tax, which amounted to RMB2,357.9 million in the first half of 2025 alone.
• Financing activities serve as the primary source of liquidity, with net cash generated from financing reaching over RMB3.3 billion in 2024 to offset operating losses.
• Cash and cash equivalents have grown significantly from RMB63.1 million in 2022 to over RMB2.5 billion by mid-2025, despite heavy operational spending.

For the six months ended June 30, 2025, our net cash used in operating activities amounted to RMB1,327.2 million, primarily attributable to the loss before tax of RMB2,357.9 million.

• The company's investing activities are primarily driven by the purchase and disposal of wealth management products and property equipment.
• A significant shift occurred in the first half of 2025, with net cash used in investing activities reaching RMB556.4 million, largely due to a RMB500 million investment in wealth management products.
• Financing activities show a massive upward trajectory, with net cash generated rising from RMB191.2 million in 2022 to over RMB3.3 billion in 2024.
• The primary source of financing is the issuance of financial instruments to investors, supplemented in 2025 by RMB700 million from convertible bonds.
• Research and development costs have seen an explosive increase, jumping from approximately RMB25.9 million in 2022 to over RMB1.4 billion in 2024.

For the six months ended June 30, 2025, our net cash generated from financing activities amounted to RMB2,165.1 million, mainly attributable to (i) proceeds from the issuance of financial instruments to investors of RMB1,625.0 million and (ii) proceeds from the issuance of convertible bonds of RMB700.0 million.

• The text provides a detailed breakdown of cash operating costs, including research and development, workforce employment, and product marketing expenses.
• Operating costs are adjusted for changes in working capital and exclude non-cash items like share-based payments and depreciation.
• The company's total indebtedness saw a massive increase from RMB 470.8 million in 2022 to over RMB 10.7 billion by October 2025.
• A significant driver of this debt is 'financial instruments issued to investors,' which grew from RMB 457.9 million to nearly RMB 9.9 billion.
• Lease liabilities, primarily for office space and computing hardware, represent another major component of the company's financial obligations.
• The financial data reflects a period of rapid scaling, with total operating costs and liabilities ballooning over a three-year track record period.

Our indebtedness during the Track Record Period primarily consisted of bank loans and lease liabilities.

• Lease liabilities experienced a massive 2,011% surge in 2023, driven by aggressive expansion into new office spaces and computing hardware.
• The company maintains a clean credit history with no defaults and holds significant liquidity through RMB6,130.5 million in unutilized bank facilities.
• Financial instruments issued to investors grew exponentially from RMB458 million to nearly RMB10 billion, reflecting heavy reliance on pre-IPO capital investments.
• Short-term convertible bonds totaling RMB830 million were issued in 2024 and 2025, all of which were converted into equity-linked instruments by May 2025.
• Bank borrowing remains relatively low compared to investor-led financing, with recent loans secured at competitive interest rates between 2.11% and 2.65%.

Our lease liabilities increased by 2,011.0% from RMB12.8 million as of December 31, 2022 to RMB270.9 million as of December 31, 2023 as we rented new premises as offices and additional computing hardware, in line with our business expansion.

• The company reports no material contingent liabilities as of June 30, 2025.
• Research and development (R&D) expenditure experienced massive growth, rising from approximately RMB 81 million in 2022 to over RMB 2.19 billion in 2024.
• R&D dominates the company's financial profile, consistently accounting for approximately 80% of total operating expenditures during the Track Record Period.
• Selling, marketing, and general administrative expenses also scaled significantly alongside R&D, though they remain secondary to technical investment.
• Capital expenditures peaked in 2023 at RMB 508.6 million, primarily driven by property and equipment purchases, before tapering off in the first half of 2025.
• Future capital expenditures are expected to be funded through a combination of existing cash, operating activities, and proceeds from the Global Offering.

During the Track Record Period, our R&D expenditure primarily consisted of R&D expenses adjusted by intangible assets related to R&D software acquired from third parties and capitalized and amortization expenses of capitalized intangible assets included in R&D expenditure.

• The company reports a significant increase in revenue growth ratio, rising from 1.2 in 2023 to 3.3 by mid-2025.
• Liquidity ratios show a downward trend, with the current and quick ratios both dropping to 0.3 as of June 30, 2025.
• Credit risk is managed through individual customer evaluations and a shift toward advance payments for cloud-based services.
• Concentration of credit risk has decreased significantly, with the top five debtors' share of receivables falling from 74.9% to 39.0%.
• The company maintains no material capital commitments or off-balance sheet arrangements as of the latest reporting period.

As of December 31, 2022, 2023 and 2024 and June 30, 2025, 74.9%, 75.3%, 50.9% and 39.0% of the total trade receivables were due from our five largest trade debtors, respectively.

• The company has not paid dividends historically and does not anticipate paying any cash dividends in the foreseeable future.
• Future dividend payments are subject to director discretion and strict PRC regulatory requirements regarding after-tax profits and statutory reserves.
• Accumulated losses and a lack of distributable reserves as of mid-2025 make it unlikely the company will be eligible to pay dividends soon.
• Related-party transactions were conducted on an arm's length basis and are not expected to distort future performance results.
• Total estimated listing expenses are approximately RMB158.5 million, with a significant portion to be charged directly to equity.
• The company intends to retain all available funds to prioritize the development and expansion of its business operations.

Investors should not purchase our ordinary shares with the expectation of receiving cash dividends.

• The company estimates legal and accounting fees for the listing at approximately RMB21.8 million, with additional expenses totaling RMB18.4 million.
• Directors confirm there have been no material adverse changes to the group's financial or trading position since June 30, 2025.
• An unaudited pro forma statement illustrates how the Global Offering would transform the group's consolidated net tangible liabilities into net tangible assets.
• The pro forma adjustment accounts for an estimated RMB3.8 billion in net proceeds and a significant RMB9.5 billion impact from the termination of financial instruments.
• Calculations are based on a fixed offer price of HK$116.20 per H Share and assume the conversion of liabilities upon the completion of the offering.

The unaudited pro forma statement of adjusted consolidated net tangible assets has been prepared for illustrative purposes only and because of its hypothetical nature, it may not give a true picture of the financial position of the Group.

• Financial instruments totaling over RMB 9.5 billion will be reclassified from liabilities to equity upon the company's listing.
• The pro forma net tangible assets calculation is based on a total of 440,230,190 shares, including those from the Global Offering.
• Calculations exclude potential shares from the Over-allotment Option or existing share incentive plans.
• Financial figures were converted from RMB to Hong Kong dollars at a specific exchange rate of 1.1016 for reporting purposes.
• The report does not account for any trading results or business transactions occurring after the June 30, 2025 cutoff date.
• The company has officially entered into cornerstone investment agreements as part of the placing process.

Upon the Listing, the redemption rights of the financial instruments will be automatically terminated, and the financial instruments issued to investors will be reclassified from liabilities to equity accordingly.

• Cornerstone Investors have committed to subscribing for H Shares totaling approximately HK$2,984.2 million at an offer price of HK$116.20 per share.
• The investment represents 68.63% of the total Offer Shares, assuming the Over-allotment Option is not exercised.
• The company intends for this placement to raise its market profile and signal investor confidence in its long-term business prospects.
• Cornerstone Investors will not receive board representation or preferential rights beyond the guaranteed allocation of shares at the final Offer Price.
• The shares acquired will rank equally with existing shares and contribute toward the company's public float requirements under Listing Rules.
• While most investors are independent third parties, three specific funds are identified as close associates of existing shareholders.

Our Company is of the view that, leveraging on the Cornerstone Investors’ investment experience and market position, the Cornerstone Placing will help to raise the profile of our Company and to signify that such Cornerstone Investors have confidence in our Company’s business and prospects.

• The text outlines strict prohibitions against side agreements between cornerstone investors and company insiders regarding the acquisition or voting of shares.
• It confirms that cornerstone investors must remain independent from the company's directors, supervisors, and substantial shareholders.
• A critical requirement is that the subscription of offer shares cannot be financed, directly or indirectly, by the company or its close associates.
• The document notes that the Stock Exchange has granted a specific waiver from strict compliance with certain regulatory requirements.
• These measures are designed to ensure transparency and prevent market manipulation during the initial public offering process.

none of the subscription of the Offer Shares by the Cornerstone Investors is directly or indirectly financed by our Company, the Directors, the Supervisor, chief executive of our Company, substantial Shareholders, existing Shareholders or any of their respective subsidiaries or their respective close associates.

• The company has obtained specific waivers from Listing Rules to allow close associates of existing shareholders to participate as cornerstone investors.
• Cornerstone investors have confirmed they possess sufficient internal resources or managed assets to fully finance their share subscriptions.
• The final number of shares allocated to cornerstone investors is subject to reallocation based on the demand levels of the Hong Kong Public Offering.
• The Sponsor-Overall Coordinator maintains absolute discretion to adjust allocations to ensure compliance with public float and placing guidelines.
• Provisions are in place for delayed delivery of shares to certain cornerstone investors to facilitate over-allocation and stabilization efforts.

Such delayed delivery arrangement is in place to facilitate the over-allocation in the International Placing.

• The document details the allocation of offer shares to various cornerstone investors, totaling over 25.6 million shares.
• JSC International Investment Fund SPC is the largest participant, contributing US$179 million for approximately 32% of the offer shares.
• The total investment from all listed cornerstone entities amounts to approximately HK$2.98 billion.
• Ownership structures reveal significant state involvement, specifically through the Beijing State-owned Assets Supervision and Administration Commission (SASAC).
• Other notable investors include WT Asset Management, GF Fund, and Taikang Life, representing a diverse mix of asset management and insurance capital.
• The shareholding percentages are calculated based on the total issued share capital immediately following the completion of the Global Offering.

JSC Management Consulting (Beijing) is controlled by Beijing Financial Holdings Group Co., Ltd., which is wholly owned by State-owned Assets Supervision and Administration Commission of the Beijing Municipal People’s Government.

• Beijing SASAC maintains significant control over Jingneng Tongxin and Beijing Energy Group through total ownership.
• JinYi Capital's participation in the Global Offering is funded by the Tsinghua University Education Foundation, a national non-public fundraising entity.
• The Tsinghua University Education Foundation operates under strict PRC regulations to preserve and enhance asset value for educational development.
• Perseverance Asset Management, a Singapore-licensed firm, manages multiple funds and has previously acted as a cornerstone investor for major companies like CATL.
• CICC Financial Trading Limited facilitates cross-border delta-one OTC equity swap transactions for Shanghai Gaoyi as part of the investment structure.

The funding of JinYi Capital Multi-Strategy Fund SPC Ltd. — Structured Credit SP Fund, which is participating in the Global Offering, are from Tsinghua University Education Foundation.

• CICC FT acts as a legal holder of Offer Shares on a non-discretionary basis to hedge OTC swap transactions for ultimate clients.
• The economic risks and returns of the shares are fully transferred to the Gaoyi Ultimate Clients, who also fully fund the swaps.
• CICC FT maintains legal title but voluntarily waives voting rights associated with the shares during the swap term per internal policy.
• A six-month lock-up period applies, after which clients may request early termination and cash settlement through share disposal.
• The ultimate clients are investment funds managed by Shanghai Gaoyi, a PRC-based asset manager focused on secondary market investments.
• The Stock Exchange granted specific consent to allow this allocation despite potential 'connected client' regulatory restrictions.

Despite that CICC FT will hold the legal title of the Offer Shares by itself, it will not exercise the voting rights attaching to the relevant Offer Shares during the terms of the Gaoyi OTC Swaps according to its internal policy.

• WT Asset Management is a Hong Kong-based firm managing approximately US$4.25 billion in assets for institutional investors like pension funds and family offices.
• The WT Funds focus on achieving absolute returns by investing in companies with significant exposure to or impact from the PRC market.
• Taikang Life, a subsidiary of Taikang Insurance Group, provides comprehensive personal security and wealth management products across diverse demographic segments.
• Taikang Insurance Group operates as a major financial conglomerate specializing in insurance, asset management, and elderly care services.
• GF Fund Management is a long-standing institution licensed to manage high-level assets including the National Social Security Fund and Basic Pension Insurance Funds.
• The listed cornerstone investors represent a mix of private asset management, insurance conglomerates, and state-linked pension fund managers.

The Funds pursue to achieve absolute return and long-term capital appreciation by investing primarily in the listed securities of companies which have great exposure or material impact by the PRC.

• GF Fund Management is a major asset manager controlled by GF Securities, which holds a 54.53% stake.
• GF Fund HK serves as the international window for its parent company, holding multiple SFC licenses for securities dealing and asset management.
• GF Fund Management and its HK subsidiary act as discretionary managers for several specific investment accounts and funds.
• 3W Fund Management, owned by Weiwei WU, focuses on absolute returns and long-term capital growth through fundamental value investing.
• RIME Capital, majority-owned by ZHUO Ying, manages the RIME Selected Shares LPF for cornerstone subscription.
• All mentioned cornerstone entities and their underlying funds are identified as independent third parties with diversified ownership.

As GF Fund Management’s window company overseas, GF Fund HK strategically connects China and the overseas market.

• Optimas Capital Limited, a Hong Kong-based investment manager owned by Ms. Cai Yun, is participating via the Optimas GCM Fund SPC.
• Luster LightTech International Limited, a subsidiary of a Shanghai-listed company, is investing as both a cornerstone investor and an existing customer of the Group.
• Luster's investment focuses on international trade in optoelectronic information and does not require additional shareholder or exchange approval.
• The cornerstone obligations are contingent upon the effectiveness of both Hong Kong and International Underwriting Agreements.
• Finalization of the investment requires the Listing Committee's approval for H Share dealings and the absence of any legal injunctions prohibiting the transaction.
• Investors must maintain the accuracy of all representations and warranties through the closing date to fulfill their contractual obligations.

Luster is one of our customers. Our Group has entered into sales contracts with Luster in connection with our provision of services to Luster in the ordinary course of business on an arm’s length basis on normal commercial terms.

• Cornerstone Investors are bound by a six-month lock-up period following the Listing Date, prohibiting the disposal of shares except under limited circumstances like transfers to wholly-owned subsidiaries.
• The company expects to receive net proceeds of approximately HK$4,173.4 million from the Global Offering, based on an offer price of HK$116.20 per share.
• A significant 70% of the proceeds is earmarked for strengthening research and development in general-purpose large AI models through 2028.
• Investment priorities include enhancing pre-trained large models, developing deep reasoning capabilities, and optimizing training infrastructure to reduce costs.
• The company plans to allocate 10% of the total proceeds specifically toward AI agents to facilitate intelligent automation and industry-specific deployment.

By these means, our models will be able to autonomously explore, reflect and operate in both online and real-world environments, enabling them to tackle complex tasks and continuously improve their performance through self-learning and reflection.

• The company plans to invest HK$417.3 million into emerging technologies to maintain its leadership in the Chinese advanced AI market.
• A significant portion of funds is dedicated to strengthening the GLM framework and building high-quality corpora for large model training.
• The strategy includes optimizing the Model-as-a-Service (MaaS) platform to offer diverse foundation models across various parameter scales.
• Investment is focused on reducing hardware barriers by engineering models compatible with consumer-grade hardware and IoT devices.
• A dedicated 5.0% of proceeds will be used to bolster sales, marketing, and brand penetration to better serve targeted client markets.

By engineering models that adapt to differing computational capabilities, we reduce hardware barriers and democratize access to advanced AI technologies.

• The company plans to allocate HK$417.3 million toward developing its business partner network and pursuing strategic investments through 2028.
• A primary goal is to integrate more computing infrastructure partners into the upstream value chain to improve model versatility.
• Investment targets include companies with mature AI products, stable operations of at least two years, and strong R&D capabilities.
• The strategy focuses on both controlling and non-controlling equity stakes in firms specializing in large models and intelligent algorithms.
• These acquisitions are intended to accelerate model training efficiency and enhance multimodal capabilities for more advanced AI solutions.

Specifically, we plan to invite more computing infrastructure partners to join our upstream value chain, further enhancing the versatility and adaptability of our models.

• The company identifies over 3,000 potential acquisition targets as of mid-2025, though no specific agreements or letters of intent have been finalized.
• Approximately 10% of the net proceeds, totaling HK$417.3 million, is earmarked for general working capital and corporate purposes.
• Unused proceeds will be held in short-term interest-bearing accounts at licensed financial institutions to ensure liquidity and compliance.
• The Hong Kong Underwriting Agreement, signed in late 2025, involves a consortium of major financial institutions led by CICC and Huatai.
• Underwriters have agreed to subscribe for shares not taken up by the public, provided specific listing approvals and conditions are met.

However, there is no guarantee that we will be able to invest in or acquire the targets that we desire.

• The Sponsor-Overall Coordinator maintains absolute discretion to terminate the Hong Kong Underwriting Agreement before the Listing Date.
• Termination can be triggered by changes in laws, regulations, or judicial interpretations across multiple jurisdictions including Hong Kong, the PRC, and the US.
• Broad economic and political shifts, such as currency devaluations or changes to the Hong Kong dollar's peg system, serve as grounds for cancellation.
• The agreement includes extensive force majeure clauses covering everything from pandemics and volcanic eruptions to civil commotion and acts of terrorism.
• Market-specific disruptions, such as trading moratoriums on major global stock exchanges or banking activity suspensions, allow for immediate withdrawal.
• The requirement to issue a prospectus supplement without prior coordinator consent provides a specific procedural trigger for termination.

The Sponsor-Overall Coordinator (for itself and on behalf of the Hong Kong Underwriters) can, in its sole and absolute discretion, by a notice in writing to us, terminate the Hong Kong Underwriting Agreement with immediate effect.

• The document outlines specific legal and regulatory triggers that allow underwriters to terminate the global offering agreement.
• Key triggers include investigations by political or regulatory bodies against company directors or senior management members.
• Economic sanctions, export controls, or the withdrawal of trading privileges in relevant jurisdictions constitute grounds for termination.
• Internal corporate changes, such as the removal of the Board chairman or directors being charged with indictable offenses, are listed as material risks.
• The Sole Sponsor maintains absolute discretion to determine if an event has a 'Material Adverse Effect' on the group's financial position or the offering's success.
• Any materialization of risks previously disclosed in the 'Risk Factors' section can lead to the cancellation of the underwriting commitment.

which, in any such case individually or in the aggregate, in the sole and absolute opinion of the Sole Sponsor and the Sponsor-Overall Coordinator... has or will or may have a material adverse effect.

• The Sole Sponsor and Underwriters reserve the right to act if statements in the prospectus or CSRC filings are found to be untrue, inaccurate, or misleading.
• Material omissions or misstatements discovered after the prospectus date that would have been relevant at the time of issuance are grounds for concern.
• Breaches of warranties or obligations by the Company or Warranting Shareholders that result in a Material Adverse Effect can trigger underwriting clauses.
• The agreement covers potential liabilities arising from indemnities or the withdrawal of the prospectus and Global Offering by the Company.
• Specific regulatory failures, such as the refusal or revocation of listing approval by the Listing Committee, are listed as critical events.
• Legal or financial instability, including winding-up orders, liquidations, or the appointment of receivers for any Group member, constitutes a material breach.

any estimate, forecast, expression of opinion, intention or expectation contained in any such Relevant Documents, was, when it was issued, or has become unfair or misleading in any respect or based on untrue, dishonest or unreasonable assumptions or given in bad faith

• The underwriting agreement may be terminated if CSRC filings are rejected, withdrawn, invalidated, or found to be non-compliant with applicable laws.
• Termination can occur if a material portion of bookbuilding orders or cornerstone investment commitments are withdrawn or fail to settle on time.
• The Company is restricted from issuing new shares or convertible securities for six months following the Listing Date, subject to specific exceptions.
• Controlling Shareholders are bound by a 12-month lock-up period during which they cannot dispose of or encumber their beneficial interests in the company's securities.
• Shareholders must immediately notify the Company if they pledge their shares as security for a loan or if they receive notice that such pledged shares will be disposed of.

Each member of the Controlling Shareholders has undertaken to the Stock Exchange and to our Company that, it shall not and shall procure that the relevant registered holder(s) will not without the prior written consent of the Stock Exchange... dispose of, nor enter into any agreement to dispose of or otherwise create any options, rights, interests or encumbrances.

• The company is required to notify the Stock Exchange and issue public announcements regarding any changes in controlling shareholder interests.
• A specific group of Key Persons, including several doctors and associated entities, have entered into irrevocable lock-up agreements.
• Key Persons are prohibited from disposing of or encumbering their shares for a period of 12 months following the Listing Date.
• Exceptions to the disposal ban include pledges for bona fide commercial loans from authorized banking institutions.
• Key Persons must immediately inform the company in writing if they pledge their shares or if they receive notice that pledged shares will be sold.
• The company commits to transparency by disclosing any such share pledges or potential disposals to the Stock Exchange and the public.

Each of the Key Persons has further irrevocably and unconditionally undertaken to us and the Stock Exchange, and shall procure its/his/her respective close associates, that within the period... ending on the date which is 12 months from the Listing Date, it/he/she will: (a) when it/he/she (or its/his/her respective close associate) pledges or charges any securities in our Company... immediately inform us in writing.

• Pathfinder Sophisticated Independent Investors (SIIs) are bound by an irrevocable six-month lock-up period following the Listing Date.
• The restrictions prohibit the disposal of securities or the creation of options and encumbrances, with limited exceptions for bona fide commercial loans.
• SIIs must immediately notify the Company in writing if they pledge their shares as security to an authorized financial institution.
• The Company is obligated to inform the Stock Exchange and issue a public announcement if any pledged shares are at risk of being disposed of by a lender.
• The Company itself is restricted from issuing or allotting new shares during the 'First Six-Month Period' without prior written consent from underwriters.

Each of the Pathfinder SIIs has irrevocably and unconditionally undertaken to us and to the Stock Exchange that except pursuant to the Global Offering, or the Over-allotment Option, it shall not... dispose of, nor enter into any agreement to dispose of or otherwise create any options, rights, interests or encumbrances.

• The text outlines strict legal restrictions on the company regarding the issuance, sale, or transfer of any equity securities during a specified lock-up period.
• Prohibitions extend beyond direct sales to include swaps, derivative arrangements, or any transaction that transfers the economic consequences of ownership.
• A secondary six-month period is defined where the company must ensure any allowed transactions do not create a 'disorderly or false market' in the shares.
• Dr. Liu, identified as the Warranting Shareholder, is legally bound to ensure the company complies with these restrictive undertakings.
• The agreement covers a wide range of financial instruments, including warrants, options, and convertible securities, whether settled in cash or stock.

we will take all reasonable steps to ensure that such transaction, offer, agreement or announcement will not create a disorderly or false market in the Shares

• The agreement prohibits the sale, mortgage, or transfer of any company shares or equity interests by specified parties for a 12-month period following the listing date.
• Restrictions extend to swaps or any financial arrangements that transfer the economic consequences of ownership to another party.
• The lock-up applies regardless of whether transactions are settled via physical delivery of shares, cash payments, or occur after the initial period.
• Parties are strictly forbidden from even announcing an intention to dispose of securities or entering into transactions with similar economic effects.
• In the event of a permitted disposal after the lock-up, the party must ensure the action does not create a 'disorderly or false market' for the company's securities.
• Any pledging or charging of shares as collateral must be immediately reported in writing to the company and the Sponsor-Overall Coordinator.

he will take all reasonable steps to ensure that such a disposal will not create a disorderly or false market in the securities of our Company

• Warranting Shareholders are permitted to acquire additional securities provided they maintain the minimum public float and adhere to existing lock-up agreements.
• Shares may be used as collateral for bona fide commercial loans from authorized financial institutions, subject to specific notification requirements.
• The Company is obligated to publicly disclose any information regarding the pledging or disposal of shares by Warranting Shareholders as required by the Stock Exchange.
• Hong Kong Underwriters currently hold no shareholding interest but may acquire H Shares through the fulfillment of their underwriting obligations.
• The International Offering is contingent upon the execution of an agreement that includes an Over-allotment Option for up to 15% of the initial Offer Shares.
• Failure to enter into or the termination of the International Underwriting Agreement will result in the cancellation of the entire Global Offering.

Potential investors should note that in the event that the International Underwriting Agreement is not entered into or terminated, the Global Offering will not proceed.

• Underwriters receive a fixed commission of 2% of the aggregate offer price, covering sub-underwriting and other internal fees.
• The company may award an additional discretionary incentive fee of up to 1% of the total offer price.
• Total estimated expenses for the global offering, including commissions and professional fees, reach approximately HK$194.4 million.
• The Sole Sponsor is entitled to a specific fee of US$500,000 for their role in the listing process.
• The company and warranting shareholders provide indemnification to underwriters against losses incurred during the performance of their duties.
• Syndicate members are permitted to engage in diverse financial activities globally that are independent of the underwriting and stabilization process.

The ratio of the Fixed Fee and the Discretionary Fee payable to all Underwriters is therefore 67:33.

• Syndicate members and their affiliates engage in diverse financial activities including trading securities, derivatives, and credit default swaps for their own accounts.
• These entities may act as agents or principals in H Share transactions, potentially providing financing to initial purchasers secured by the shares themselves.
• Market-making and liquidity provision requirements on various stock exchanges often necessitate hedging activities that directly involve the H Shares.
• Such trading and hedging activities can occur globally and may result in syndicate members holding conflicting long and short positions.
• The collective actions of these financial institutions may negatively impact the trading price, liquidity, and volatility of the H Shares.
• Strict regulations prohibit syndicate members, excluding the Stabilizing Manager, from conducting transactions intended to artificially maintain the market price.

Those activities may require hedging activity by those entities involving, directly or indirectly, the buying and selling of the H Shares, which may have a negative impact on the trading price of the H Shares.

• Syndicate members must strictly adhere to market misconduct laws, including prohibitions on insider dealing and price rigging.
• The Sole Sponsor and its affiliates maintain ongoing commercial relationships with the company, providing investment banking and financing services.
• The Global Offering is divided into a Hong Kong Public Offering and an International Offering, totaling over 37 million H Shares.
• Investors are restricted to participating in either the Hong Kong Public Offering or the International Offering, but not both.
• The total Offer Shares represent approximately 8.50% of the company's enlarged issued share capital, potentially rising to 9.65% if the Over-allotment Option is exercised.

The Syndicate Members must comply with all applicable laws and regulations, including the market misconduct provisions of the SFO, including the provisions prohibiting insider dealing, false trading, price rigging and stock market manipulation.

• The company is initially offering 1,871,000 H Shares to the public in Hong Kong, representing 5% of the total Global Offering.
• The offering is open to both the general public and professional investors, including brokers, fund managers, and corporate entities.
• Allocation of shares is based on the volume of valid applications and may involve balloting if the offering is oversubscribed.
• Available shares are divided into two equal pools: Pool A for applications of HK$5 million or less, and Pool B for applications exceeding HK$5 million.
• To prevent market concentration, multiple applications or any single application for more than 50% of the public offering shares will be rejected.

Such allocation could, where appropriate, consist of balloting, which could mean that some applicants may receive a higher allocation than others who have applied for the same number of Hong Kong Offer Shares.

• The distribution of shares is divided between the Hong Kong Public Offering and the International Offering.
• A mandatory clawback mechanism is triggered based on specific demand levels in the Hong Kong market.
• The Listing Rules dictate how shares are reallocated to increase the Hong Kong portion when demand is high.
• Reallocation thresholds are specifically defined for demand levels exceeding 10 times and 50 times the initial supply.
• These structural requirements ensure a fair distribution of shares to public investors during high-demand scenarios.

Paragraph 4.2 of Practice Note 18 of the Listing Rules (as modified by Rule 18C.09 of the Listing Rules) requires a clawback mechanism to be put in place.

• The Sponsor-Overall Coordinator maintains discretionary power to reallocate shares between the Hong Kong Public Offering and the International Offering based on demand.
• Mandatory reallocation triggers can increase the Hong Kong Public Offering to 10% or 20% of total shares depending on the level of oversubscription.
• Specific provisions allow for a discretionary increase of up to 3,742,000 shares for Hong Kong applicants if the International Offering is undersubscribed or Hong Kong demand is moderate.
• If the Hong Kong Public Offering is undersubscribed, the remaining shares may be shifted back to the International Offering at the coordinator's discretion.
• The entire Global Offering faces cancellation if both segments are undersubscribed, unless underwriters step in to procure the remaining balance.
• Final details regarding the distribution and reallocation of shares must be publicly disclosed in a results announcement by early January 2026.

Where the International Offer Shares are not fully subscribed, if the Hong Kong Offer Shares are also not fully subscribed, the Global Offering will not proceed unless the Underwriters would subscribe or procure subscribers.

• Applicants must confirm they are not participating in both the Hong Kong Public Offering and the International Offering simultaneously.
• Breaching the non-duplication undertaking may result in the rejection of the applicant's submission.
• The initial cost for a single board lot of 100 shares is set at HK$11,737.19, including various levies and fees.
• The International Offering comprises 95% of the total initial Offer Shares, totaling 35,548,500 H Shares.
• The International Offering represents approximately 8.07% of the company's total shares in issue post-offering, assuming no over-allotment.

Such applicant’s application under International Offering is liable to be rejected if such undertaking and/or confirmation is/are breached and/or untrue.

• The International Offering targets institutional and professional investors outside the United States, utilizing a book-building process to establish a stable shareholder base.
• Allocation decisions prioritize investors based on demand timing, asset size, and the likelihood of holding shares long-term rather than immediate selling.
• Listing Rules require that at least 50% of the total Global Offering be allocated to independent price-setting investors to ensure market integrity.
• Mechanisms are in place to prevent 'double-dipping' by excluding investors from the International Offering if they have also applied through the Hong Kong Public Offering.
• The company may grant an Over-allotment Option allowing the issuance of up to 15% additional shares to cover excess demand within 30 days of the offering.
• Total share counts remain subject to change through clawback arrangements and the potential exercise of the Over-allotment Option.

Such allocation is intended to result in a distribution of the H Shares on a basis which would lead to the establishment of a solid professional and institutional shareholder base to the benefit of our Company and the Shareholders as a whole.

• Underwriters may intervene in the secondary market to prevent the share price from falling below the initial offer price.
• The Stabilizing Manager has the discretion to over-allocate shares or establish short positions to support market stability.
• In the Hong Kong market, any stabilizing bids or purchases are strictly prohibited from exceeding the original Offer Price.
• Stabilization activities are temporary and must conclude within 30 days of the Hong Kong Public Offering application deadline.
• Investors are warned that once the stabilization period ends, the lack of artificial support may cause the share price to decline.
• The liquidation of long positions held by the Stabilizing Manager can itself create downward pressure on the market price.

However, there is no obligation on the Stabilizing Manager (or its affiliates or any person acting for it) to conduct any such stabilizing action.

• The Stabilizing Manager may execute transactions at or below the Offer Price to stabilize the share value after listing.
• Stabilization is supported by delayed delivery arrangements for up to 15% of the initial Offer Shares.
• The exercise of the Over-allotment Option is strictly dependent on investor agreement to these delayed delivery terms.
• The fixed Offer Price is set at HK$116.20 per share, with specific additional levies and fees for public applicants.
• The Sponsor-Overall Coordinator maintains the authority to reduce the Offer Price or the number of shares based on investor interest.

If no investor in the International Offering agrees to the delayed delivery arrangements, no stabilizing actions will be undertaken by the Stabilizing Manager and the Over-allotment Option will not be exercised.

• The company reserves the right to reduce the number of Offer Shares or the Offer Price prior to the final application deadline.
• Any significant reduction triggers a cancellation and relaunch of the Global Offering via a supplemental prospectus on the FINI platform.
• The Sponsor-Overall Coordinator maintains discretionary power to reallocate shares between the Hong Kong Public and International Offerings.
• Public notices regarding price or share reductions may be published as late as the morning of the final application day.
• The entire offering remains conditional upon the Hong Kong Stock Exchange granting formal listing approval and permission to deal in H Shares.

The Global Offering will be canceled and subsequently relaunched on FINI pursuant to the supplemental prospectus

• The Global Offering is contingent upon the execution of the International Underwriting Agreement and both offerings remaining unconditional.
• If specified conditions are not met or waived by the deadline, the offering will lapse and all application monies will be returned without interest.
• H Share certificates are scheduled for issuance on January 7, 2026, but only become valid title evidence on the morning of January 8, 2026.
• Trading of H Shares on the Hong Kong Stock Exchange is expected to commence at 9:00 a.m. on January 8, 2026, under stock code 2513.
• The application process is fully electronic, with strict eligibility criteria regarding age, residency, and legal status in the U.S. and mainland China.

H Share certificates for the Offer Shares are expected to be issued on Wednesday, January 7, 2026, but they will only become valid evidence of title at 8:00 a.m. on Thursday, January 8, 2026.

• The text outlines strict eligibility criteria, excluding existing shareholders, directors, and connected persons from participating in the public offering.
• The application window is strictly defined, running from December 30, 2025, to noon on January 5, 2026.
• Two primary application channels are provided: the HK eIPO White Form for physical certificates and the HKSCC EIPO channel for electronic deposits.
• Applicants are warned that both digital platforms are subject to capacity limitations and potential service interruptions during peak times.
• The legal framework establishes that completing payment constitutes a binding application and a declaration that no multiple applications were made.

The HK eIPO White Form service and the HKSCC EIPO channel are facilities subject to capacity limitations and potential service interruptions and you are advised not to wait until the last day.

• Investors can authorize brokers or custodians to apply for Hong Kong Offer Shares through the HKSCC EIPO channel.
• HKSCC Nominees acts as a legal intermediary, assuming no liability for actions taken on behalf of the applicant.
• Strict anti-duplication rules apply, where multiple applications by a single person across different channels lead to total rejection.
• Individual applicants must provide specific identity documentation, prioritizing HKID cards over national IDs or passports.
• Corporate applicants are required to provide LEI registration or certificates of incorporation to validate their legal status.
• Joint applications are limited to a maximum of four individuals and require a valid Hong Kong address and contact details.

If you are suspected of making more than one application through the HK eIPO White Form service or any other channel, all of your applications are liable to be rejected.

• Applicants must provide both English and Chinese names if they possess both, with strict adherence to HKID and LEI identification standards.
• Trustees and investment funds are required to provide specific client identification data (CID) linked to their asset management companies or trading accounts.
• Joint account applications are strictly limited to a maximum of four holders in accordance with FINI market practices.
• Nominee applications must disclose full identity details for all beneficial owners to avoid the application being treated as the nominee's own.
• Unlisted companies under an applicant's statutory control are treated as the applicant's own benefit, requiring full disclosure of control structures.
• Applications through the HKSCC EIPO channel may require pre-funding as determined by brokers or custodians based on local regulations.

If you do not include this information, the application will be treated as being made for your benefit.

• The document outlines the specific payment amounts required for different tiers of Hong Kong Offer Shares, ranging from 100 to 935,500 shares.
• Payments are inclusive of brokerage fees, SFC transaction levies, Stock Exchange trading fees, and AFRC transaction levies.
• Applicants using the HK eIPO White Form service must pay the full amount at the time of application to be considered.
• Strict prohibitions are in place against multiple applications by the same individual or for the same benefit across different channels.
• The H Share Registrar utilizes a system to identify and reject suspected multiple applications based on names and identification numbers.
• The maximum number of shares an individual can apply for is capped at 50% of the initial Hong Kong Offer Shares.

If you are suspected of submitting or cause to submit more than one application, all of your applications will be rejected.

• Applicants authorize agents to execute all necessary documents for registering and depositing shares into the CCASS system.
• By applying, investors confirm they have read the prospectus and agree to be bound by all stated terms and procedures.
• Liability for the company and its underwriters is strictly limited to information and representations contained within the official prospectus.
• Investors must consent to the disclosure of personal data to regulatory bodies, including the SFC and the Stock Exchange.
• Applications are generally irrevocable once accepted and cannot be rescinded due to innocent misrepresentation.
• The agreement mandates compliance with HKSCC General Rules and Operational Procedures for all electronic application instructions.

Agree (without prejudice to any other rights which you may have once your application has been accepted) that you will not rescind it because of an innocent misrepresentation.

• Applicants must confirm they are not being financed by the company's directors or substantial shareholders to ensure independence.
• The agreement stipulates that all applications and resulting contracts are governed strictly by the laws of Hong Kong.
• Applicants must warrant that they are not 'U.S. Investors' and that the shares have not been registered under the U.S. Securities Act.
• The terms strictly prohibit multiple applications for the same beneficiary to prevent market manipulation or unfair allocation.
• Applicants acknowledge that providing false declarations is a criminal offense that may lead to prosecution.
• The document requires confirmation that the applicant has not participated in the International Offering portion of the global share sale.

confirm that you understand that we and the Overall Coordinators will rely on your declarations and representations in deciding whether or not to allocate any Hong Kong Offer Shares to you and that you may be prosecuted for making a false declaration;

• Applicants must warrant that their purchase complies with all applicable international laws to avoid legal breaches for the issuing entity.
• Allotment results are accessible through designated websites, telephone inquiry lines, and the Stock Exchange's official portal starting January 7, 2026.
• HKSCC Participants have early access to review allotment results via the FINI system to report any discrepancies before the general public announcement.
• The company and its coordinators maintain absolute discretion to reject any application or accept only a portion of it without providing justification.
• Share allocations become void if the Stock Exchange fails to grant listing permission within a specific three-to-six-week window following the application close.
• Multiple or suspected duplicate applications are strictly prohibited and serve as grounds for immediate disqualification from the offering.

We, the Overall Coordinators, the H Share Registrar and their respective agents and nominees have full discretion to reject or accept any application, or to accept only part of any application, without giving any reasons.

• Applications for Hong Kong Offer Shares may be rejected due to incomplete instructions, payment errors, or potential legal violations.
• HKSCC Participants must maintain sufficient funds on deposit, but a risk of money settlement failure remains during the balloting process.
• In cases of extreme settlement failure by a broker or custodian, allotted shares may be reallocated to the International Offering, leaving the applicant with nothing.
• The company and relevant registrars disclaim all liability for non-allocation resulting from third-party money settlement failures.
• H Share certificates only become valid evidence of title on a specific future date, provided the Global Offering becomes unconditional.
• Specific collection procedures exist for large-scale applications, requiring in-person pickup or electronic deposit into CCASS accounts.

In the extreme case, you will not be allocated any Hong Kong Offer Shares due to the money settlement failure by such HKSCC Participant.

• Applicants must provide valid identification to the H Share Registrar to collect share certificates in person.
• Uncollected certificates or those for applications under 500,000 shares are mailed via ordinary post at the applicant's risk.
• Refunds for surplus application monies are processed via e-Auto Refund for single accounts or physical checks for multiple accounts.
• The refund timeline is primarily set for January 8, 2026, though specific timing depends on individual broker arrangements.
• Severe weather protocols, including tropical cyclones or black rainstorms, can delay the opening and closing of application lists.
• Contingency plans are in place to ensure share certificate delivery to HKSCC even during periods of extreme weather conditions.

If you do not collect your H Share certificate(s) personally within the time above, it/they will be sent to the address specified in your application instructions by ordinary post at your own risk

• Postponements in the application process may lead to delays in the official listing date of H Shares.
• Specific contingency plans are in place for severe weather signals, including modified delivery schedules for physical share certificates.
• H Shares will be admitted into the Central Clearing and Settlement System (CCASS) upon approval from the Stock Exchange and HKSCC.
• Settlement between Exchange Participants must occur on the second settlement day following any trading day.
• The collection and use of personal data from applicants are strictly governed by the Personal Data (Privacy) Ordinance of Hong Kong.

If any of Severe Weather Signal is hoisted on Wednesday, January 7, 2026, the H Share Registrar will make appropriate arrangements for the delivery of the H Share certificates to the CCASS Depository’s service counter so that they would be available for trading on Thursday, January 8, 2026.

• Applicants must provide accurate personal data to ensure the successful processing of Hong Kong Offer Shares and the dispatch of share certificates.
• Inaccurate or missing data can lead to the rejection of applications, delays in share transfers, or the inability to receive entitled services.
• Personal data is utilized for verifying identities, identifying duplicate applications, and facilitating the balloting process for share allocation.
• The company uses collected information to manage shareholder registers, establish benefit entitlements like dividends, and distribute corporate communications.
• Data may be shared with third-party agents, financial advisers, and regulatory bodies both within and outside Hong Kong to fulfill operational obligations.
• Confidentiality is maintained, but disclosure is permitted to entities like HKSCC Nominees and service providers necessary for business operations.

Failure to supply the requested data or supplying inaccurate data may result in your application for Hong Kong Offer Shares being rejected, or in the delay or the inability of our Company or the H Share Registrar to effect transfers or otherwise render their services.

• The document outlines the legal framework for sharing personal data with regulatory bodies like the SFC and Stock Exchange.
• Personal data of share applicants is retained only as long as necessary and is governed by the Personal Data (Privacy) Ordinance of Hong Kong.
• Shareholders maintain the right to access, verify, and correct their personal data held by the Company or the H Share Registrar.
• KPMG provides a comprehensive Accountants' Report on the historical financial information of Knowledge Atlas Technology Joint Stock Company Limited.
• The financial audit covers a Track Record Period spanning from December 2022 through June 2025.
• Company directors are legally responsible for ensuring the Historical Financial Information provides a true and fair view, free from material misstatement.

Personal data which is no longer required will be destroyed or dealt with in accordance with the Personal Data (Privacy) Ordinance (Chapter 486 of the Laws of Hong Kong).

• The reporting accountants are responsible for expressing an opinion on the historical financial information provided in the investment circular.
• Work was conducted under the Hong Kong Standard on Investment Circular Reporting Engagements 200 to ensure freedom from material misstatement.
• Procedures included assessing risks of fraud or error and evaluating the appropriateness of accounting policies and estimates made by directors.
• Internal controls were considered to design appropriate procedures, though no opinion was expressed on the effectiveness of those controls.
• The accountants concluded that the financial information provides a true and fair view of the Group's financial position from 2022 through mid-2025.
• A review was also conducted on the stub period corresponding financial information for the first half of 2024.

The procedures selected depend on the reporting accountants’ judgement, including the assessment of risks of material misstatement of the Historical Financial Information, whether due to fraud or error.

• KPMG provides a limited assurance review of the company's stub period financial information, noting that a review is significantly less in scope than a full audit.
• The company reported no adjustments to the underlying financial statements and confirmed that no dividends were paid during the track record period.
• Revenue shows significant growth, rising from approximately RMB 57.4 million in 2022 to over RMB 312.4 million in 2024.
• Research and development expenses have escalated dramatically, reaching over RMB 2.19 billion in 2024, contributing to substantial operational losses.
• The consolidated statements reveal deepening net losses, driven by high selling, administrative, and R&D costs alongside changes in the fair value of financial instruments.

A review is substantially less in scope than an audit conducted in accordance with Hong Kong Standards on Auditing and consequently does not enable us to obtain assurance that we would become aware of all significant matters that might be identified in an audit.

• The company has experienced rapidly accelerating annual losses, growing from approximately RMB 143 million in 2022 to nearly RMB 3 billion by the end of 2024.
• A significant financial burden is driven by 'Financial instruments issued to investors,' which ballooned from RMB 458 million to over RMB 9.5 billion in just three and a half years.
• The consolidated statement of financial position reveals a severe net liability position, reaching a deficit of over RMB 6.1 billion by June 2025.
• Despite heavy losses, the company maintains a substantial cash position, with cash at bank and on hand increasing to over RMB 2.5 billion by mid-2025.
• The loss per share has worsened significantly over the reported periods, dropping from RMB 7.80 in 2022 to RMB 62.27 for the first half of 2025 alone.

The accompanying notes form part of this Historical Financial Information.

• The company exhibits a significant and growing net liability position, with the total deficit increasing from approximately 180 million RMB in 2022 to nearly 5.9 billion RMB by mid-2025.
• A primary driver of the financial strain is the 'Financial instruments issued to investors,' which surged from 458 million RMB to over 9.5 billion RMB during the reported period.
• Current liabilities consistently exceed current assets, resulting in a widening net current liability gap that reached nearly 7.9 billion RMB by June 2025.
• Despite the heavy losses, the company has significantly increased its investments in subsidiaries and property and equipment, indicating aggressive expansion or infrastructure building.
• The consolidated statements of changes in equity reveal that accumulated losses are the primary factor eroding the capital base, far outweighing capital contributions.

The accompanying notes form part of this Historical Financial Information.

• The financial data reveals a significant and accelerating trend of accumulated losses, growing from approximately 210 million RMB in early 2023 to nearly 4 billion RMB by mid-2025.
• A major structural shift occurred in 2025 with the conversion into a joint stock limited liability company, involving a massive transfer of 4.24 billion RMB from capital reserves to offset accumulated losses.
• Equity-settled share-based transactions increased dramatically over the period, jumping from 5.5 million RMB in 2023 to over 161 million RMB in the first half of 2025 alone.
• The company's total equity position has deteriorated into a deep deficit, reaching a negative 3.95 billion RMB by June 30, 2025.
• Despite heavy losses, the company continued to receive capital contributions from equity shareholders and non-controlling interests to bolster paid-in capital.

Conversion into a joint stock limited liability company 30(b) - (4,243,292) - - - 4,243,292 - - -

• The company experienced a significant widening of accumulated losses, reaching over RMB 2.2 billion by mid-2024.
• Total equity shows a substantial deficit, indicating that liabilities significantly outweigh assets on the balance sheet.
• Operating activities consistently consume large amounts of cash, with net cash used in operations exceeding RMB 1.3 billion in the first half of 2025 alone.
• Losses before taxation have accelerated sharply year-over-year, jumping from RMB 143 million in 2022 to nearly RMB 3 billion in 2024.
• Significant non-cash adjustments include depreciation, changes in the carrying amounts of financial instruments issued to investors, and share-based payments.

Balance at 30 June 2024 (unaudited): (2,247,895) Total equity - deficit.

• The company, formerly known as Beijing Zhipu Huazhang Technology, transitioned to a joint stock limited liability company in March 2025.
• Financing activities show a massive influx of capital, with proceeds from financial instruments rising from 208 million RMB in 2022 to over 3 billion RMB in 2024.
• Investing activities reflect significant capital expenditure on property and equipment, peaking at over 506 million RMB in 2023.
• The Group is heavily active in wealth management products, frequently moving hundreds of millions of RMB in and out of these short-term investments.
• Cash and cash equivalents have grown exponentially from approximately 63 million RMB at the start of 2022 to over 2.5 billion RMB by mid-2025.
• The Group's primary business focus is the provision of large model-related services within the People’s Republic of China.

The Company and its subsidiaries (together, the “Group”) are principally engaged in the provision of large model-related services in the PRC.

• The document details the corporate structure of a group as of June 30, 2025, listing numerous wholly-owned subsidiaries across China, Hong Kong, Singapore, and the British Virgin Islands.
• A significant majority of the subsidiaries are dedicated to the 'provision of large model-related services,' indicating a heavy strategic focus on artificial intelligence and LLM development.
• The group has undergone rapid expansion between 2023 and 2024, establishing multiple new entities in major Chinese tech hubs like Hangzhou, Shanghai, and Tianjin.
• Capitalization varies significantly across the entities, ranging from nominal amounts for offshore investment holdings to RMB 950,000,000 for the Tianjin-based technology subsidiary.
• The auditing history shows a transition for many entities from RSM China CPA LLP to Beijing Chengju Certified Public Accountants between 2022 and 2024.

Beijing Knowledge Xingyao Technology Co., Ltd. (北京智譜興曜科技有限公司) (i) The PRC 24 September 2024 RMB300,000,000 100% 100% - Provision of large model-related services

• The document details a rapid series of subsidiary incorporations between late 2024 and mid-2025 across China, the UK, and Malaysia.
• All listed subsidiaries are primarily focused on the provision of large model-related services, indicating a heavy investment in AI infrastructure.
• The Group maintains 100% ownership in the vast majority of these entities, with registered capital reaching as high as RMB 450 million for specific regional branches.
• Despite the aggressive expansion, the Group reported a significant net loss of over RMB 2.3 billion for the first half of 2025.
• Financial health indicators show substantial net liabilities of RMB 6.15 billion, largely driven by financial instruments issued to investors.
• The financial statements are prepared in accordance with IFRS Accounting Standards and comply with Hong Kong Stock Exchange disclosure rules.

For the six months ended 30 June 2025, the Group incurred net loss of RMB2,357,852,000 and as at 30 June 2025, the Group recorded net liabilities of RMB6,150,841,000.

• The Group maintains a going concern status based on cash flow forecasts through June 2026 and over RMB 2.3 billion in unutilized banking facilities.
• Financial instruments issued to investors, totaling over RMB 9.5 billion, are expected to convert into equity upon the Company's stock exchange listing.
• Historical Financial Information is prepared primarily on a historical cost basis, with specific exceptions for investments and convertible bonds measured at fair value.
• Management utilizes estimates and judgments based on historical experience to determine asset and liability carrying values where direct data is unavailable.
• The Group defines subsidiary control as having exposure to variable returns and the power to affect those returns through involvement with the entity.
• Intra-group balances and unrealized gains or losses from internal transactions are eliminated during the consolidation of financial statements.

The directors of the Company are of the opinion that no payment is expected for the settlement of the liabilities arising from financial instruments issued to investors within twelve months from the date of this report.

• Non-controlling interests (NCI) can be measured at fair value or as a proportionate share of net identifiable assets.
• Changes in subsidiary interests that do not result in a loss of control are treated strictly as equity transactions.
• Losing control of a subsidiary triggers the derecognition of assets and liabilities, with any retained interest remeasured at fair value.
• Investments in associates are managed via the equity method, where the Group recognizes its share of profits or losses until significant influence ceases.
• Goodwill is recorded at cost less accumulated impairment and must undergo annual impairment testing.
• Securities investments are initially stated at fair value, with subsequent accounting determined by their specific classification.

When the Group loses control of a subsidiary, it derecognises the assets and liabilities of the subsidiary, and any related NCI and other components of equity.

• Non-equity investments are measured at fair value through profit or loss (FVPL), with all value changes and interest recognized immediately.
• Equity investments are generally classified as FVPL unless an irrevocable election is made to designate them as fair value through other comprehensive income (FVOCI).
• Property and equipment are recorded at cost and depreciated using the straight-line method, with electronic equipment typically lasting three to five years.
• Research expenditure is expensed as incurred, while development costs are only capitalized if technical and commercial feasibility can be demonstrated.
• Intangible assets like patents and software are amortized over periods ranging from one to eight years, subject to annual reviews of their useful lives.
• Lease assessments focus on whether a contract conveys the right to control an identified asset and obtain substantially all its economic benefits.

Development expenditure is capitalised only if the expenditure can be measured reliably, the product or process is technically and commercially feasible, future economic benefits are probable and the Group intends to and has sufficient resources to complete development and to use or sell the resulting asset.

• The Group treats lease and non-lease components as a single lease component for accounting purposes.
• Right-of-use assets and lease liabilities are recognized at commencement, excluding short-term or low-value leases.
• Lease liabilities are initially measured at the present value of future payments using implicit or incremental borrowing rates.
• Right-of-use assets include initial liability amounts, direct costs, and restoration estimates, subsequently depreciated over the lease term.
• Lease liabilities are remeasured following changes in indices, rates, or lease modifications, with adjustments applied to the asset's carrying amount.
• The Group establishes loss allowances for expected credit losses on financial assets measured at amortized cost and contract assets.

When the lease liability is remeasured in this way, a corresponding adjustment is made to the carrying amount of the right-of-use asset, or is recorded in profit or loss if the carrying amount of the right-of-use asset has been reduced to zero.

• Expected Credit Losses (ECLs) are calculated as probability-weighted estimates of cash shortfalls discounted to their present value.
• The Group distinguishes between 12-month ECLs for low-risk assets and lifetime ECLs for assets with significantly increased credit risk.
• Trade receivables and contract assets are subject to a simplified approach where loss allowances are always measured at lifetime ECLs.
• Credit risk assessment incorporates both historical experience and forward-looking information to identify significant increases in default probability.
• Financial assets are deemed credit-impaired when specific detrimental events occur, such as debtor bankruptcy or breach of contract.
• Assets are written off entirely when there is no realistic prospect of recovery, such as a lack of debtor assets or income sources.

The gross carrying amount of a financial asset or contract asset is written off to the extent that there is no realistic prospect of recovery.

• The Group evaluates non-financial assets for impairment at each reporting date, estimating recoverable amounts if indicators of value loss exist.
• Recoverable amounts are determined by identifying the higher of an asset's 'value in use' through discounted cash flows or its fair value minus disposal costs.
• Impairment losses are prioritized to reduce goodwill first, and while general asset impairments can be reversed, goodwill impairments are permanent.
• Inventories are strictly measured at the lower of their transaction cost or their net realisable value in the ordinary course of business.
• Contract costs are capitalized only if they are expected to be recovered and relate directly to future revenue generation, otherwise they are expensed immediately.
• Contract assets represent revenue recognized before an unconditional right to payment exists, while contract liabilities track prepayments for future goods or services.

An impairment loss in respect of goodwill is not reversed.

• Trade receivables are recognized when the Group has an unconditional right to consideration, typically measured at transaction price or fair value.
• Cash and cash equivalents are defined as highly liquid, short-term investments with insignificant risk of value changes and maturities under three months.
• Financial liabilities arise from contracts requiring the Group to purchase its own equity instruments, even if the obligation is conditional.
• Convertible bonds are designated as financial liabilities measured at fair value through profit or loss (FVPL).
• Short-term employee benefits and retirement plan contributions are expensed as services are rendered, provided a legal or constructive obligation exists.

A contract that contains an obligation for the Company or the Group to purchase its own equity instruments for cash or another financial asset gives rise to a financial liability even if the Company’s or the Group’s obligation to purchase is conditional on the counterparty exercising its right to redeem.

• Equity-settled share-based payments are valued using the binomial model and expensed over the vesting period.
• Expense adjustments for equity awards are based on the likelihood of service conditions being met by the vesting date.
• Modifications to equity-settled payments require additional expense recognition if they increase the fair value for the employee.
• Termination benefits are recorded as expenses when the offer becomes irrevocable or during restructuring.
• Income tax expense includes current and deferred tax, generally recognized in profit or loss unless linked to equity or business combinations.
• Deferred tax is not recognized for specific temporary differences, such as the initial recognition of goodwill or certain subsidiary investments.

The amount recognised as an expense is adjusted to reflect the equity awards for which the related service conditions are expected to be met, such that the amount ultimately recognised is based on the equity awards that meet the related service conditions at the vesting date.

• The Group accounts for deferred tax assets and liabilities separately for leases, basing asset recognition on the probability of future taxable profits.
• Provisions for warranties and onerous contracts are calculated using discounted future cash flows and historical probability data.
• Contingent liabilities are disclosed when economic outflows are possible but not probable, or when amounts cannot be reliably estimated.
• Revenue is recognized only when control of a product or service is transferred to the customer, excluding third-party taxes.
• The Group distinguishes between acting as a principal or agent based on its ability to direct the use of and obtain benefits from services.
• Core business operations focus on large model-related services through both on-premise and cloud-based deployment.

Control refers to the Group’s ability to direct the use of and obtain substantially all of the remaining benefits from the product or service.

• Revenue for on-site large model deployment is recognized upon delivery and customer acceptance at a designated location.
• Cloud-based services generate revenue either ratably over a subscription term or based on specific customer usage metrics.
• Interest income is calculated using the effective interest method, with specific adjustments for credit-impaired financial assets.
• Government grants are treated as deferred income and recognized as expense reductions in the periods the related costs are incurred.
• Foreign currency transactions are translated at transaction-date rates, while foreign operations are translated into RMB for presentation.
• The document defines related parties based on control, significant influence, or membership in key management personnel.

For usage-based contract, revenue is recognised based on the customer’s utilisation of the resources when the services are rendered to the customers.

• The text defines the complex criteria for identifying related parties, including joint ventures, associates, and post-employment benefit plans.
• Segment reporting is determined by how senior executive management allocates resources and assesses performance across business lines and geographies.
• Material operating segments can only be aggregated if they share similar economic characteristics, products, and regulatory environments.
• Management must exercise significant judgement when estimating the recoverable amount of non-current non-financial assets to determine impairment.
• Impairment assessments rely on discounting expected future cash flows, which are sensitive to revenue projections and operating cost assumptions.
• The Group's primary revenue source is identified as providing large model-related services within the PRC.

In determining the value in use, expected future cash flows generated by the asset are discounted to their present value, which requires significant judgement relating to the level of revenue and amount of operating costs.

• The Group's revenue is primarily driven by on-premise deployment, which significantly outpaces cloud-based deployment in total earnings.
• Financial data shows a rapid growth trajectory, with total revenue increasing from approximately 57.4 million RMB in 2022 to over 312.4 million RMB in 2024.
• Revenue recognition occurs predominantly at a specific point in time, though the 'over time' recognition category has seen steady growth.
• The company maintains a diverse but shifting customer base, with different major clients exceeding the 10% revenue threshold in different fiscal years.
• Segment reporting is divided into on-premise and cloud-based services, with gross profit serving as the primary measure for performance assessment.

On-premise deployment: this segment develops and provides customised large model-related services according to the customers’ specific instructions and needs at the customers’ infrastructure.

• The Group operates two primary reportable segments: on-premise deployment and cloud-based deployment, with the former generating the vast majority of revenue and gross profit.
• Total revenue experienced explosive growth from RMB 57.4 million in 2022 to over RMB 312 million in 2024, though the cloud-based segment recorded a small gross loss in the first half of 2025.
• Despite rising revenues, the Group's consolidated loss before taxation widened significantly, reaching nearly RMB 2.96 billion in 2024, primarily driven by massive research and development spending.
• Research and development expenses are the largest cost driver, ballooning from RMB 84.4 million in 2022 to over RMB 2.19 billion in 2024.
• Financial results are heavily impacted by non-cash items, including changes in the carrying amounts of financial instruments issued to investors and fair value adjustments.
• The Group's operations are entirely concentrated within the PRC, with all revenue derived from domestic customers and all non-current assets located there.

Consolidated loss before taxation (143,650) (787,957) (2,958,007) (1,235,551) (2,357,852)

• Total staff costs experienced a massive surge from RMB90.7 million in 2022 to over RMB573 million by 2024, driven largely by salaries and equity-settled compensation.
• Equity-settled share-based compensation expenses saw a dramatic spike in the first half of 2025, reaching RMB158.8 million compared to just RMB4.2 million in the same period of 2024.
• The Group consistently reported significant losses before taxation, with the deficit widening from RMB143.7 million in 2022 to nearly RMB3 billion in 2024.
• Tax benefits from research and development deductions provided a significant offset, totaling RMB305.4 million in 2024 alone.
• The Group maintains retirement benefit schemes including local government-managed plans in Mainland China and the Mandatory Provident Fund (MPF) in Hong Kong.
• Despite the heavy losses, the Group has not recognized the tax effect of unused tax losses, which grew to RMB510.5 million by mid-2025.

The Group has no further obligation for payment of other retirement benefits beyond the above contributions.

• The PRC Corporate Income Tax laws increased the additional deduction for qualified R&D expenses from 75% to 100% effective October 1, 2022.
• Certain financial instrument adjustments and share-based payment expenses are classified as non-deductible for tax purposes under PRC regulations.
• Executive compensation structures for directors and supervisors include salaries, discretionary bonuses, retirement contributions, and share-based payments.
• Total director and supervisor emoluments showed a significant upward trend, increasing from approximately 5 million RMB in 2022 to over 11.5 million RMB in 2024.
• A massive spike in share-based payments is recorded for Dr. Liu Debing in the first half of 2025, reaching over 27 million RMB.

An additional 100% of qualified research and development expenses incurred is allowed to be deducted from taxable income under the PRC Corporate Income Tax laws and regulations after 1 October 2022.

• The report provides a granular breakdown of director and supervisor compensation, including salaries, discretionary bonuses, and retirement contributions.
• A significant portion of total compensation for top executives is derived from share-based payments, which are measured according to specific accounting policies.
• There was notable turnover in the leadership team during 2025, with several directors resigning or being re-appointed within a three-month window.
• The composition of the 'five highest paid individuals' shifted over the track record period, moving from a majority of directors to a majority of non-directors.
• Non-director compensation saw a massive spike in the first half of 2025, primarily driven by over 51 million RMB in equity-settled share-based compensation expenses.

These represent the estimated value of equity awards granted to the directors under the Group’s equity award schemes.

• The company reports significant and increasing net losses, with the loss attributable to ordinary shareholders rising from RMB 103.9 million in 2022 to nearly RMB 1.4 billion in 2024.
• Basic loss per share has escalated sharply over the track record period, moving from RMB 7.80 in 2022 to RMB 87.20 in 2024.
• The calculation of weighted average shares includes a complex adjustment for the company's conversion into a joint stock company in March 2025.
• Financial instruments issued to investors significantly impact the loss allocation, accounting for over RMB 1.5 billion in adjustments in 2024 alone.
• Diluted loss per share remains identical to basic loss per share because including convertible bonds and other instruments would be anti-dilutive.
• Initial property and equipment costs were primarily driven by electronic equipment and right-of-use assets, totaling RMB 59 million by the end of 2022.

During the Track Record Period, financial instruments issued to investors and convertible bonds were not included in the calculation of diluted loss per share as their inclusion would have been anti-dilutive.

• The report details a significant expansion in asset costs, with total carrying amounts growing from approximately RMB 41 million in 2022 to over RMB 772 million by mid-2025.
• A major financing arrangement involved transferring legal ownership of electronic equipment to a third party while retaining use through a four-year lease.
• Right-of-use assets became a dominant portion of the balance sheet, reflecting a strategic shift toward leased rather than owned infrastructure.
• Depreciation charges accelerated sharply over the period, peaking at over RMB 270 million for the year 2024.
• The financial data tracks three primary categories: electronic equipment, leasehold improvements, and right-of-use assets across both Group and Company levels.

The Group entered into a financing arrangement with a third party, under which the Group transferred the legal ownership of the electronic equipment to this third party while continue to use these equipment under lease for a term of four years.

• The financial report details a significant increase in right-of-use assets, particularly electronic equipment, which grew from zero in 2022 to over 403 million RMB by mid-2025.
• A specific financing arrangement is highlighted where the Group transferred legal ownership of electronic equipment to a third party while maintaining use through a four-year lease.
• Depreciation charges for right-of-use assets saw a sharp rise, jumping from approximately 12 million RMB in 2022 to over 176 million RMB in 2024.
• Intangible assets, including patents and software, expanded through both direct additions and business combinations, reaching a cost basis of over 66 million RMB by the end of 2023.
• Impairment tests conducted at the Cash Generating Unit (CGU) level confirmed that recoverable amounts exceeded carrying amounts, resulting in no recognized impairment losses.

The Group entered into a financing arrangement with a third party, under which the Group transferred the legal ownership of the electronic equipment to this third party while continue to use these equipment under lease for a term of four years.

• The report details the financial tracking of patents and software assets from 2022 through mid-2025, showing significant growth in software additions.
• A major acquisition of Beijing Lingxin Intelligent in 2023 introduced 39.379 million RMB in goodwill to the Group's balance sheet.
• Management utilizes an extended eight-year forecast period for cash flow projections due to the high-growth nature of the underlying technology.
• Key valuation assumptions include aggressive annual revenue growth rates peaking at 35% and steady gross profit margins between 54% and 58%.
• The pre-tax discount rate applied to the Beijing Lingxin Intelligent unit has gradually decreased from 19% in 2023 to 17% by mid-2025.

Management adopted a forecast period of longer than five years in view that the business is still under significant growth and will require additional time for the underlying technology to reach stable status.

• The Group conducted impairment tests on the Beijing Lingxin Intelligent CGU, concluding that no goodwill impairment was necessary through mid-2025.
• Sensitivity analysis indicates that a pre-tax discount rate increase of only 2% to 3% would eliminate the remaining financial headroom for the CGU.
• Investments in subsidiaries saw exponential growth, rising from RMB 46.4 million in 2022 to over RMB 1.19 billion by June 2025.
• The Group expanded its interests in associates, notably acquiring significant stakes in computing and technical service firms like Beijing Xinglian and Beijing Doushen Zhichuang.
• Significant influence is maintained over Beijing Shudao despite a low 7.31% equity interest, due to board representation and policy-making participation.

The following tables set out the hypothetical change to pre-tax discount rate that would have removed the remaining headroom.

• The Group maintains a 35.68% effective interest in its primary associates, though the carrying amount of these investments decreased significantly between late 2024 and mid-2025.
• Aggregate data for immaterial associates shows a trend of increasing net losses, growing from 453,000 RMB in 2023 to nearly 5.9 million RMB by June 2025.
• Wealth management product investments saw a massive surge in the first half of 2025, jumping from zero to over 500 million RMB.
• The company is strategically invested in unlisted AI-related industries, categorizing these as equity securities measured at fair value through profit or loss.
• Significant prepayments are being made for computing service fees to secure future processing capacity, reflecting the infrastructure-heavy nature of their operations.
• Input VAT deductibles and contract fulfillment costs have scaled rapidly, indicating a period of high operational expenditure and expanding project volume.

The prepayments for computing service fees represented payments made to suppliers to secure these suppliers’ computing services for a certain period, and will be deducted by subsequent utilisation.

• The Group's trade receivables experienced a massive surge, growing from approximately RMB 6.6 million in 2022 to over RMB 171 million by mid-2025.
• Contract fulfilment costs for the Company rose significantly, reaching RMB 66.5 million in 2025, indicating a substantial increase in operational scale or project volume.
• Loss allowances for trade receivables have escalated sharply, jumping from RMB 29,000 in 2022 to nearly RMB 25.9 million in 2025, reflecting heightened credit risk.
• A major divestment occurred in November 2024, where the Group sold unlisted equity investments to an associate, Beijing Xinglian, for a total of RMB 202.5 million.
• The ageing analysis reveals a shift in credit collection, with a growing portion of receivables now falling into the 6 months to 1 year category as of June 2025.
• The Company maintains significant prepayments for computing service fees, which peaked at RMB 127.4 million in 2024 before declining in 2025.

The consideration was determined based on arm’s length negotiation and was with reference to the most recent transaction price of the equity interests in those unlisted entities.

• Contract assets from on-premise deployment contracts showed significant growth, rising from RMB 405,000 in 2022 to over RMB 17.8 million by mid-2025.
• The Group's cash and cash equivalents experienced a massive surge, increasing from approximately RMB 219 million in 2022 to over RMB 2.5 billion by June 2025.
• Restricted cash balances are primarily tied to bidding deposits and performance guarantees, with fund remittances subject to PRC foreign exchange controls.
• Time deposits placed in PRC financial institutions carry interest rates between 3.00% and 3.70%, with shifting classifications between current and non-current portions.
• Financing activities in 2022 were heavily driven by the issuance of financial instruments to investors, totaling RMB 208 million in proceeds.

Remittance of funds out of the PRC is subject to relevant rules and regulations of foreign exchange control.

• The report details significant capital inflows from the issuance of financial instruments to investors, totaling over 3 billion RMB in the 2024 period.
• Lease liabilities show substantial growth, with a net increase of 547,773,000 RMB recorded in the final reporting segment.
• The company diversified its debt profile by introducing bank loans and convertible bonds, adding 137,166,000 RMB and 130,000,000 RMB respectively.
• Complex accounting adjustments are present, including reclassifications between other payables, other receivables, and financial instruments issued to investors.
• Finance costs and interest elements of lease rentals represent a consistent cash outflow, with interest on leases reaching 30,931,000 RMB in the latest period.

Proceeds from the issuance of financial instruments to investors - 3,019,587.

• The report details significant capital movements, including RMB 1.625 billion in proceeds from financial instruments issued to investors during the first half of 2025.
• A major reclassification occurred where RMB 833.5 million was moved from convertible bonds to financial instruments issued to investors.
• Total liabilities and financial obligations grew substantially, reaching over RMB 10.3 billion by June 30, 2025, compared to approximately RMB 7.6 billion at the start of the year.
• Lease-related cash outflows show a dramatic upward trend, increasing from RMB 14.3 million in 2022 to an annualized projection far exceeding the RMB 187.4 million spent in 2024.
• Financing activities were the primary driver of cash flow changes, characterized by high-volume issuance of instruments and convertible bonds offset by transaction costs and lease payments.

Reclassification from the convertible bond to the financial instruments issued to investors.

• The Group's total trade and other payables saw a massive surge from approximately RMB 25.8 million in 2022 to over RMB 838 million by mid-2025.
• Computing service fees represent the largest single liability component, ballooning from RMB 1.5 million to nearly RMB 495.3 million over the reporting period.
• The vast majority of trade payables are current, with the aging analysis showing most invoices are settled within a three-month window.
• Contract liabilities, representing receipts in advance from customers, stabilized at approximately RMB 75 million by 2025 after significant growth in earlier years.
• The Group introduced significant bank loans and lease liabilities starting in 2023, with lease obligations peaking at over RMB 671 million in 2024.
• All trade and other payables are expected to be settled within one year or are repayable on demand, indicating a high requirement for short-term liquidity.

All of the trade and other payables are expected to be settled within one year or are repayable on demand.

• The Group's lease liabilities show a significant upward trajectory, growing from approximately 12.8 million RMB in 2022 to over 578 million RMB by mid-2025.
• Financial instruments issued to investors represent a massive liability on the balance sheet, ballooning from 458 million RMB to over 9.5 billion RMB during the Track Record Period.
• Investors hold redemption rights that can be triggered if the Company fails to complete an initial public offering (IPO) by February 2028.
• The potential redemption cost is calculated as the original principal investment plus an accrued interest rate of 12% per annum.
• The Group also managed convertible bonds totaling 830 million RMB, which were ultimately converted into financial instruments issued to investors by June 2025.
• Redemption rights and their associated financial liabilities are slated to terminate automatically upon the successful listing of the Company’s shares on a recognized stock exchange.

The Company and these investors have also entered into equity holders’ agreements pursuant to which the Company has granted rights for these investors to require the Company to redeem the invested amounts if certain triggering events cannot be met, including an initial public offering of the Company’s shares by February 2028.

• The Group issued convertible bonds totaling RMB830 million in 2024 and 2025, which were later converted into financial instruments in May 2025.
• Fair value measurements for these bonds utilized an equity allocation model with a high expected volatility of 59.64%.
• Deferred tax assets and liabilities were primarily driven by right-of-use assets and lease liabilities, maintaining a net-zero balance through offsetting reconciliations.
• The Group has opted not to recognize significant deferred tax assets for cumulative tax losses, which escalated to over RMB8.3 billion by mid-2025.
• Management determined it is not probable that future taxable profits will be available to utilize these massive tax losses within the allowed carry-forward periods.
• Equity awards have been granted since 2021 to directors and employees to incentivize sustainable development and business performance.

The Group has not recognised deferred tax assets in respect of cumulative tax losses and deductible temporary difference of RMB8,343,905,000 as at 30 June 2025 respectively, as it is not probable that future taxable profits against which the losses can be utilised will be available.

• The company transitioned from equity awards to restricted shares in June 2025 following its conversion into a joint stock limited liability company.
• Equity awards were issued based on both service and performance conditions, with an arbitrary initial value of RMB1 per unit prior to conversion.
• The fair value of equity awards saw significant growth, rising from RMB28.12 in 2022 to nearly RMB300 by mid-2025.
• A binomial model was utilized to estimate fair values, incorporating variables such as expected volatility of comparable companies and risk-free interest rates from the People’s Bank of the PRC.
• The replacement of all outstanding equity awards with 2,016,000 restricted shares in 2025 is classified as a modification of the original equity incentive plan.
• By June 30, 2025, the total number of outstanding restricted shares reached 6,620,000 with a fair value per ordinary share of RMB302.73.

The Company has arbitrarily set RMB1 per equity unit for the purposes of granting these equity awards before the Company converted into a joint stock limited liability company.

• The report details the reconciliation of equity components including paid-in capital, reserves, and accumulated losses from 2022 through mid-2025.
• The Group experienced significant financial pressure, with accumulated losses ballooning from RMB 64.7 million in early 2022 to over RMB 1.8 billion by June 2025.
• Total equity transitioned from a positive balance of RMB 4.89 million to a substantial deficit of nearly RMB 5.9 billion over the Track Record Period.
• Share-based payment reserves saw a dramatic increase, rising from RMB 300,000 to over RMB 192 million, reflecting heavy reliance on equity-settled transactions.
• A major structural change occurred in 2025 with the conversion of the entity into a joint stock limited liability company, involving a massive transfer from capital reserves to offset accumulated losses.

The fair values of restricted shares will vary if different variables and assumptions are adopted.

• The Group underwent a significant structural transformation, converting from a private entity into a joint stock limited liability company on March 26, 2025.
• Paid-in capital saw substantial growth from 12.8 million RMB in 2022 to over 36.2 million RMB prior to the conversion, driven by investor contributions and reserve transfers.
• Upon conversion, the company's registered capital was initially divided into shares with a nominal value of RMB 1.00 each.
• A planned ten-for-one share split is scheduled to occur immediately prior to the Global Offering, reducing the nominal value per share to RMB 0.10.
• The company's capital reserves are defined by the delta between net considerations received and the nominal share capital issued during various funding rounds.
• Despite the capital growth, the Group declared no dividends during the Track Record Period, with past distributions not serving as indicators for future policy.

Immediately prior to the Global Offering, the ordinary shares of the Company will be split on a one for ten basis, and the registered share capital of our Company will be RMB40,281,069.

• The Group prioritizes maintaining a 'going concern' status by balancing shareholder returns with the security of a sound capital position.
• Financial risk management focuses on mitigating exposure to credit, liquidity, interest rate, and currency risks inherent in business operations.
• Credit risk is primarily linked to trade receivables and contract assets, though risk from bank deposits is considered low due to high counterparty credit standings.
• The Group performs individual credit evaluations on customers and generally requires advance payments for cloud-based subscription services.
• Concentration of credit risk has shifted significantly, with the top five debtors' share of trade receivables dropping from 75% in 2022 to 39% by mid-2025.
• Loss allowances are calculated using Expected Credit Loss (ECL) models, which show a sharp increase in expected loss rates for older receivables.

The Group actively and regularly reviews and manages its capital structure to maintain a balance between the higher shareholder returns that might be possible with higher levels of borrowings and the advantages and security afforded by a sound capital position.

• The report details expected loss rates for contract assets and receivables, showing a sharp increase in risk as debt ages, reaching over 83% for items outstanding for 2 to 3 years.
• Impairment losses have escalated dramatically over the track record period, rising from a mere RMB34,000 in 2022 to over RMB47 million by mid-2025.
• Specific debtors in financial difficulty were identified, leading to significant loss allowances for other receivables to account for increased credit risk.
• The Group maintains a liquidity risk policy focused on monitoring cash reserves and securing committed funding lines from financial institutions.
• Financial liability tables reveal a massive expansion in trade payables and lease liabilities, with total carrying amounts growing from RMB32 million to over RMB1.3 billion in less than three years.

The Group has identified certain debtors were in financial difficulties and the related credit risks have increased significantly.

• The Group's financial liabilities include trade payables, lease liabilities, and bank loans, with a total carrying amount of approximately RMB 1.55 billion.
• Liquidity risk is further influenced by financial instruments issued to investors and convertible bonds, which have specific payment terms.
• Fair value measurements are categorized into a three-level hierarchy based on the observability of inputs, ranging from quoted market prices to unobservable data.
• The Group's Level 3 assets, including equity securities and wealth management products, rely on significant unobservable inputs for valuation.
• Valuation techniques for Level 3 instruments include the market approach, discounted cash flow method, and equity allocation models.
• During the reporting period, there were no transfers between the different levels of the fair value hierarchy.

Unobservable inputs are inputs for which market data are not available.

• The Group utilizes fair value through profit or loss (FVPL) for non-listed equity investments, primarily valued based on recent financing rounds or market approaches.
• Sensitivity analysis reveals that a 1% change in expected returns on wealth management products could impact pre-tax losses by up to RMB5,000,000 as of mid-2025.
• Interest rate risk is managed across a portfolio of lease liabilities and bank loans, with fixed-rate borrowings increasing significantly from 2022 to 2025.
• The Group executed a strategic business combination by acquiring the remaining 94.98% of Beijing Lingxin Intelligent for RMB73,005,000.
• The acquisition of Beijing Lingxin Intelligent, an AI services provider, resulted in the recognition of RMB39,379,000 in goodwill.
• Financial instruments carried at amortized cost are reported to have carrying amounts that do not materially differ from their fair values.

The acquisition was made as part of the Group’s strategy to expand its market share of artificial intelligence in the PRC.

• The acquisition of Beijing Lingxin Intelligent resulted in a net cash outflow of RMB64.1 million, despite the subsidiary contributing a net loss of RMB5.9 million in 2023.
• Key management personnel remuneration saw a massive spike in the first half of 2025, reaching RMB43.4 million, primarily driven by equity-settled share-based compensation.
• The Group maintains complex financial relationships with several equity holders and associates, including significant trade receivables and payables totaling over RMB116 million by mid-2025.
• A major non-trade transaction occurred in 2024 involving the divestment of equity interests in unlisted companies valued at RMB206 million.
• The company identifies a specific group of individuals and partnerships, including several doctors and technology centers, as the ultimate controlling parties.

If the acquisition had occurred on 1 January 2023, management estimates that consolidated revenue would have increased by RMB3,889,000, and consolidated loss for the year would have increased by RMB13,910,000.

• The IASB has issued several new standards and amendments, including IFRS 18 and IFRS 19, that are not yet effective for the current Track Record Period.
• IFRS 18 will replace IAS 1 in 2027, introducing a mandatory classification of income and expenses into five distinct categories in the profit or loss statement.
• The Group anticipates that most upcoming accounting developments will not have a significant impact on its historical financial information.
• While IFRS 18 will change the presentation and disclosure of financial statements, it is not expected to alter the Group's underlying financial performance or position.
• The report confirms that no material subsequent events occurred between June 30, 2025, and the date of the report's issuance.

IFRS 18 will replace IAS 1 Presentation of financial statements and aims to improve the transparency and comparability of information about an entity’s financial statements.

• The document outlines the unaudited pro forma statement of adjusted net tangible assets for the Group as of June 30, 2025.
• It illustrates how the Global Offering and the termination of investor financial instruments would transform a consolidated net tangible liability into a positive asset position.
• A significant accounting shift occurs as RMB 9.56 billion in financial instruments are reclassified from liabilities to equity upon the company's listing.
• The estimated net proceeds from the offering of 37,419,500 H shares are calculated based on an indicative price of HK$116.20 per share.
• The resulting pro forma adjusted net tangible assets per share are estimated at RMB 16.19 or HK$ 17.85.
• The statement is purely illustrative and hypothetical, cautioned not to represent the actual future financial position of the Group.

Upon the Listing, the redemption rights of the financial instruments will be automatically terminated, and the financial instruments issued to investors will be reclassified from liabilities to equity accordingly.

• The document outlines the compilation of unaudited pro forma financial information for Knowledge Atlas Technology Joint Stock Company Limited.
• KPMG provides an independent assurance report regarding the impact of the proposed Global Offering on the company's net tangible assets.
• Financial figures are converted from Renminbi to Hong Kong dollars at a specific rate, though this does not guarantee future exchangeability.
• The pro forma data is strictly illustrative, simulating the company's financial position as if the offering had occurred on June 30, 2025.
• The directors are responsible for compliance with Listing Rules and Accounting Guideline 7 in preparing these financial statements.

No representation is made that Renminbi amount have been, could have been or may be converted to Hong Kong dollars, or vice versa, at that rate or at any other rate.

• The reporting accountants conducted their engagement under HKSAE 3420 to ensure the pro forma information aligns with Listing Rules.
• The primary purpose of the pro forma data is to illustrate the hypothetical impact of significant events as if they had occurred at an earlier date.
• Accountants explicitly state they are not responsible for updating historical financial reports or performing new audits on the underlying data.
• The report provides no assurance that the actual outcome of transactions as of June 30, 2025, would match the presented illustrations.
• The procedures followed do not comply with United States auditing standards (PCAOB) and should not be relied upon as such.
• The accountants offer no opinion on the reasonableness or actual application of the net proceeds from the share issuance.

Accordingly, we do not provide any assurance that the actual outcome of events or transactions as 30 June 2025 would have been as presented.

• The document outlines the taxation of income and capital gains for H Share holders under current PRC laws and practices.
• Individual investors are generally subject to a flat 20% income tax on dividends paid by PRC enterprises.
• Preferential tax treaty rates may allow for a reduced withholding rate of 10% for individual holders from certain jurisdictions.
• Domestic non-foreign-invested enterprises issuing shares in Hong Kong are permitted to withhold individual income tax at a standard 10% rate during distribution.
• Tax liabilities vary significantly based on the investor's residency and the specific tax treaties between their home country and the PRC.
• Investors are strongly advised to consult personal tax advisers as the summary does not account for individual circumstances or future legal changes.

The discussion does not deal with all possible tax consequences relating to an investment in the H Shares, nor does it take into account the specific circumstances of any particular investor, some of which may be subject to special regulations.

• Non-resident enterprises are generally subject to a 10% withholding tax on income sourced from within the PRC.
• The 10% tax rate specifically applies to dividends distributed by PRC-resident enterprises to overseas H-share and B-share holders.
• Tax withholding is performed at the source, meaning the payer must deduct the tax before distributing funds to the non-resident entity.
• Tax treaties or specific international agreements may modify these standard withholding rates for eligible entities.
• Under the Mainland-Hong Kong tax arrangement, the dividend tax rate can be reduced to 5% if a Hong Kong resident holds at least 25% of the PRC company's equity.

The aforesaid income tax payable for non-resident enterprises is deducted at source, where the payer of the income is required to withhold the income tax from the amount to be paid to the non-resident enterprise.

• New criteria effective December 2019 restrict treaty benefits if the main purpose of an arrangement is deemed to be tax avoidance.
• Non-resident investors from jurisdictions with double taxation treaties may qualify for reduced enterprise income tax on dividends.
• Preferential tax rates are not automatic and require non-PRC resident enterprises to apply for refunds subject to tax authority approval.
• The transfer of marketable securities is generally subject to a 6% Value-Added Tax (VAT) on the gain for general or foreign taxpayers.
• Individual investors are specifically exempt from VAT when transferring financial products.
• Local surcharges typically amounting to 12% of the VAT payable are applied, though certain exemptions exist for overseas entities.

The treaty benefits under the criteria shall not be granted in the circumstance where relevant gains, after taking into account all relevant facts and conditions, are reasonably deemed to be one of the main purposes for the arrangement or transactions.

• Overseas entities are currently exempt from urban maintenance and construction taxes and education surcharges on VAT paid for intangible asset sales since September 2021.
• There is significant regulatory ambiguity regarding whether non-PRC resident enterprises must pay VAT on the disposal of H shares, which could impact investment value.
• Individual income tax on the transfer of listed company stocks has historically been temporarily exempted, though recent law revisions lack explicit clarity on the continuation of this policy.
• The current tax exemptions for individuals primarily focus on shares traded on the Shanghai and Shenzhen exchanges, leaving the status of overseas-listed PRC shares uncertain.
• Non-resident enterprises are generally subject to a 10% withholding tax on income derived from the sale of equity interests in PRC-resident enterprises.
• Tax liabilities for non-resident enterprises may be mitigated through specific double taxation avoidance treaties or international agreements.

If relevant tariffs are imposed in the future, the investment value of such holders in H shares may be materially and adversely affected.

• The Stamp Tax Law of the PRC applies to taxable certificates and securities transactions conducted both within and outside the territory if used domestically.
• As of the latest reporting date, there is currently no estate duty levied within the People's Republic of China.
• The State Administration of Foreign Exchange (SAFE) manages all foreign exchange matters under the authorization of the People's Bank of China.
• Current account international payments and transfers are generally unrestricted and do not require SAFE approval.
• Capital account transactions remain subject to strict foreign exchange regulations and existing restrictions on convertibility.

As at the Latest Practicable Date, no estate duty is levied within the PRC.

• The People's Bank of China transitioned from a fixed USD peg to a managed floating exchange rate system based on market supply and a basket of currencies.
• Refinements in 2014 and 2015 introduced a weighted average calculation for central parity quotations based on market maker input and previous closing rates.
• The 2008 Foreign Exchange Control Regulations shifted the focus toward balancing capital inflows and outflows while granting the State emergency intervention powers.
• PRC enterprises can now conduct current account transactions and profit distributions in foreign exchange through designated banks without prior administrative approval.
• The regulatory framework has evolved to grant the State Administration of Foreign Exchange (SAFE) extensive supervisory authority over international balances of payment.

In the event that international balance of payment suffer or may suffer a material misbalance, or the national economy encounters or may encounter a severe crisis, the State may adopt necessary safeguard or control measures.

• The State Council removed the requirement for SAFE approval regarding the remittance and settlement of proceeds from overseas listings into domestic RMB accounts.
• Domestic companies must register their overseas listings with local SAFE branches within 15 business days of completing an issuance.
• Proceeds from overseas listings can be repatriated to China or kept abroad, provided the usage aligns with public disclosure documents.
• Foreign exchange registration for direct investments is now handled directly by banks rather than SAFE, shifting the agency to a supervisory role.
• Eligible enterprises may utilize capital account income for domestic payments without item-by-item authenticity verification prior to each transaction.
• The current tentative percentage for foreign exchange settlement of capital account earnings is set at 100%, subject to future adjustments by SAFE.

Eligible enterprises are allowed to make domestic payments by using their capital, foreign credits and the income under capital accounts of overseas listing, without providing materials to the bank in advance for authenticity verification on an item-by-item basis.

• The appendix provides a summary of PRC laws and regulations relevant to the company's operations, excluding taxation and foreign exchange.
• The PRC legal system is primarily based on the Constitution and consists of written laws, administrative regulations, and international treaties.
• Court judgments in the PRC do not serve as legally binding precedents but are used for judicial reference and guidance.
• The National People's Congress (NPC) and its Standing Committee hold the primary legislative power to formulate and amend basic state laws.
• The State Council acts as the highest administrative organ, with the authority to establish regulations based on existing constitutional law.
• Local governments and people's congresses have the power to create regional regulations provided they do not conflict with national laws.

Court Judgments do not constitute legally binding precedents, although they may be used for the purposes of judicial reference and guidance.

• The PRC Constitution maintains supreme legal authority, and no other laws or regulations may contravene its provisions.
• A strict hierarchy exists where national laws outrank administrative regulations, which in turn outrank local regulations and rules.
• Local governments and specific ministries have the authority to formulate rules within their jurisdictions provided they align with superior national mandates.
• The National People's Congress (NPC) and its Standing Committee possess the power to annul any regulations that conflict with the Constitution or higher laws.
• Higher-level government bodies, such as the State Council, are empowered to alter or annul inappropriate rules enacted by lower-level departments or local governments.
• The power to interpret laws is formally vested in the Standing Committee of the National People's Congress (SCNPC).

The PRC Constitution has supreme legal authority and no laws, administrative regulations, local regulations, autonomous regulations or separate regulations may contravene the PRC Constitution.

• The 1981 Resolution on Strengthening Legal Interpretation divides interpretive authority among the SCNPC, the Supreme People's Court, and the Supreme People's Procuratorate.
• The SCNPC acts as the final arbiter when the Supreme People's Court and the Supreme People's Procuratorate have fundamental differences in legal interpretation.
• Administrative and regional authorities retain the power to interpret the specific regulations and local laws they have promulgated.
• The PRC judicial hierarchy consists of the Supreme People's Court, local courts at three levels, and specialized courts for military, maritime, and intellectual property matters.
• The Supreme People's Procuratorate and higher-level procuratorates are empowered to supervise the legally effective judgments and rulings of the courts.

If there are differences in principle in the interpretation of the Supreme People’s court and the Supreme People’s Procuratorate, they shall be submitted to the SCNPC for interpretation or decision.

• The Chinese judicial system operates on a two-tier appellate structure where second-instance rulings are generally final and binding.
• Judicial supervision procedures allow for retrials if errors are discovered in final judgments by higher courts or presiding judges.
• The Civil Procedure Law governs jurisdiction, typically favoring the defendant's domicile unless parties agree on a venue with a direct connection to the dispute.
• Foreign litigants are granted reciprocal rights and obligations but must engage PRC-qualified lawyers for legal representation in domestic courts.
• International judicial assistance is based on treaties and reciprocity, provided such cooperation does not violate PRC sovereignty or public interest.
• Court judgments and arbitration awards must be performed by all parties, with a two-year window for applying for compulsory enforcement.

A PRC court shall not accommodate any request made by a foreign court which will result in the violation of sovereignty, security or public interests of the PRC.

• The PRC court system can mandatorily enforce judgments if parties fail to comply within a stipulated period.
• Procedures exist for the cross-border recognition and enforcement of judgments based on international treaties or the principle of reciprocity.
• PRC courts may refuse to enforce foreign judgments if they are deemed to violate national sovereignty, security, or public interest.
• Joint stock limited companies seeking overseas listings are primarily governed by the PRC Company Law and CSRC Trial Administrative Measures.
• The liability of a joint stock limited company is limited to its total assets, while shareholder liability is limited to their subscribed shares.
• Companies must operate according to administrative regulations and may invest in other entities without assuming joint liability for their debts.

Likewise, if the PRC has entered into either a treaty relating to judicial enforcement with the relevant foreign country or according to the principle of reciprocity, a foreign judgement or ruling may also be recognized and enforced in accordance with the PRC enforcement procedures by a PRC court unless the people’s court considers that the recognition or enforcement of such judgement or ruling would violate the basic legal principles of the PRC, its sovereignty or national security, or would not be in the public interest.

• Companies in the PRC can be established through either promotion or flotation, requiring between one and 200 promoters.
• At least half of the company's promoters must be residents within the People's Republic of China.
• In promotion-based incorporation, shares cannot be offered to external parties until the registered capital is fully paid up by the promoters.
• For flotation-based incorporation, promoters are required to subscribe to at least 35% of the total shares unless otherwise regulated.
• Subscribers have the right to demand a refund with interest if shares are not fully subscribed or if the inauguration meeting is delayed.
• Legal person status is only achieved once the business license is issued following registration with the administration for market regulation.

Where the shares issued are not fully subscribed for within the offer period stipulated in the share offering prospectus, or where the promoter fails to convene an inauguration meeting within thirty days of the subscription capital for the shares issued being fully paid up, the subscribers may demand that the promoters refund the subscription capital so paid together with the interest calculated at bank rates of a deposit for the same period.

• Promoters can contribute capital via currency or non-monetary assets like intellectual property, provided they undergo strict valuation to prevent over-valuation.
• The PRC Company Law mandates that all shares of the same class must carry equal rights and be issued at the same price and conditions.
• Share offering prices are strictly regulated to be equal to or greater than the par value, never below it.
• Companies can increase share capital through public offerings, private placements, bonus shares, or the conversion of reserve funds.
• Domestic companies seeking to offer shares overseas must file with the CSRC within three business days of their application submission.
• Specific classified shares with varying voting or distribution rights are permitted but restricted for companies making public offerings.

The share offering price may be equal to or greater than the par value of the share, but may not be less than the par value.

• Companies must follow a strict five-step legal procedure to reduce registered capital, including preparing balance sheets and notifying creditors.
• Creditors maintain the right to demand debt repayment or guarantees when a company decides to reduce its registered capital.
• Share repurchases are generally prohibited except under specific conditions such as mergers, employee stock plans, or protecting shareholder value.
• Approval for share repurchases varies by circumstance, requiring either a shareholder resolution or a two-thirds board majority.
• Strict statutory timelines exist for the disposal of repurchased shares, ranging from ten days to three years depending on the reason for acquisition.
• Listed companies are granted specific flexibility to repurchase shares to maintain corporate value and protect equity interests.

The creditors of the company may within the statutory time limit require the company to repay its debts or provide guarantees for covering the debts.

• Share transfers must be conducted through established stock exchanges or as mandated by the State Council, requiring formal endorsement and registration.
• Specific 'blackout periods' prohibit share register changes 20 days before a general meeting or five days before dividend record dates.
• Strict lock-up periods apply to pre-IPO shares, which cannot be traded for one year following the company's listing.
• Corporate insiders face annual transfer caps of 25% of their holdings and are barred from selling for six months after leaving their positions.
• Shareholders possess fundamental rights to participate in decision-making, receive dividends, and inspect corporate records and financial reports.
• Legal recourse is available to shareholders, allowing them to petition courts to revoke resolutions that violate laws or the company's articles of association.

During their terms of office, they may transfer no more than 25% of the total number of shares they hold in the company per annum.

• The shareholders’ meeting serves as the primary organ of authority within a company under PRC Company Law.
• Key powers include electing or removing directors and supervisors and determining their remuneration packages.
• Shareholders hold the exclusive right to approve major financial decisions such as profit distributions, capital changes, and bond issuances.
• The meeting is responsible for fundamental corporate structural changes, including mergers, divisions, dissolutions, and amendments to the articles of association.
• Annual meetings are mandatory, but extraordinary meetings must be triggered by specific crises, such as significant capital loss or a shortage of directors.
• Shareholders holding at least 10% of shares have the legal right to request an extraordinary meeting independently of the board.

The shareholders’ meeting is the organ of authority of the company, which exercises its powers in accordance with the PRC Company Law.

• The board of directors holds primary responsibility for convening and presiding over shareholders' meetings, with a clear hierarchy of succession for the chairperson.
• If the board of directors fails to act, the board of supervisors is mandated to convene the meeting to ensure corporate continuity.
• Shareholders holding at least 10% of shares for 90 consecutive days have the legal right to unilaterally convene a meeting if both boards fail to do so.
• Specific notice periods are required by law, with twenty days' notice for regular meetings and fifteen days for extraordinary sessions.
• The PRC Company Law does not define a specific quorum requirement for shareholders' meetings, but establishes a one-vote-per-share rule for those present.

If the board of supervisors fails to convene and preside over such meeting, shareholders individually or in aggregate holding 10% or more of the company’s shares for ninety days or more consecutively may unilaterally convene and preside over such meeting.

• Accumulative voting systems allow shareholders to consolidate votes for specific directors or supervisors, potentially empowering minority interests.
• Standard resolutions require a simple majority, while critical structural changes like mergers or capital adjustments necessitate a two-thirds majority.
• The PRC Company Law mandates that boards with 300 or more employees must include a staff representative unless one is already on the board of supervisors.
• Directors serve terms of up to three years but must continue their duties past expiration if a successor has not been duly elected to maintain a quorum.
• The board of directors holds the primary authority to formulate profit distribution plans, investment proposals, and corporate restructuring strategies.

Under the accumulative voting system, each share shall be entitled to the number of votes equivalent to the number of directors or supervisors to be elected at the shareholders’ meeting, and shareholders may consolidate their votes for one or more directors or supervisors when casting a vote.

• The board of directors holds the authority to appoint or dismiss senior management, including the manager and financial officer, and determine their compensation.
• Board meetings must occur at least twice annually, requiring a quorum of more than half the directors and a simple majority for passing resolutions.
• Directors are personally liable for compensating the company if they approve resolutions that violate laws or regulations resulting in serious losses.
• A director can be exempt from liability only if they formally recorded their objection to a harmful resolution in the meeting minutes.
• Specific legal disqualifications prevent individuals with criminal records for economic crimes or those responsible for past corporate insolvencies from serving as directors.

However, if it can be proved that a director expressly objected to the resolution when the resolution was voted on, and that such objection was recorded in the minutes of the meeting, such director shall be relieved from that liability.

• Individuals with significant unpaid debts or legal enforcement records are legally barred from serving as company directors.
• The board of directors must elect a chairman through a majority vote, with clear succession protocols if the chairman is incapacitated.
• A board of supervisors must be established, consisting of shareholder representatives and at least one-third staff representatives.
• Staff representatives on the supervisory board must be democratically elected by the company's employees.
• Supervisors serve three-year terms and are tasked with auditing financial positions and monitoring the conduct of senior management.
• The board of supervisors holds the power to propose the removal of directors who violate laws or act against the company's interests.

Where a company elects or appoints a director to which any of the above circumstances applies, such election or appointment shall be null and void.

• The board of supervisors holds the authority to investigate company irregularities and initiate legal actions against directors or senior management.
• Joint stock limited companies may substitute a supervisory committee with an audit committee composed primarily of independent directors.
• Audit committees must approve critical financial decisions, including the appointment of accounting firms and the disclosure of financial reports, before board resolution.
• Company managers are appointed by the board and may attend board meetings but lack voting rights unless they also serve as directors.
• Senior management, including financial officers and board secretaries, are legally bound by duties of faithfulness and diligence toward the company.
• Strict prohibitions are in place to prevent the misappropriation of funds, acceptance of bribes, or unauthorized disclosure of confidential company information.

The board of supervisors may investigate any irregularities identified in the operation of the company and, when necessary, may engage an accounting firm to assist its work at the cost of the company.

• Directors and senior management must report and seek formal approval for any direct or indirect contracts or transactions involving the company.
• The duty of loyalty extends to close relatives and controlled enterprises, preventing them from bypassing conflict-of-interest disclosures.
• Corporate leaders are prohibited from usurping business opportunities or engaging in competing businesses without explicit shareholder or board consent.
• Any income generated through unauthorized self-dealing or competitive activities must be returned to the company as a penalty.
• Fiduciaries are legally liable for compensation if their breach of duty, laws, or articles of association results in financial loss to the company.
• The duty of diligence requires officers to ensure all business activities remain within the legal scope of the company's license and treat shareholders equally.

Income generated by directors or senior management in violation of aforementioned shall be returned to the company.

• Companies must establish accounting systems and undergo annual audits in accordance with State Council regulations.
• Financial reports must be accessible to shareholders 20 days before annual meetings, and public companies must publish these reports.
• Statutory common reserve funds must be maintained at 10% of after-tax profits until they reach 50% of the registered capital.
• Profits must first cover previous losses and reserve fund allocations before being distributed to shareholders based on shareholding proportions.
• The capital reserve fund, derived from share premiums, can be used to cover losses or expand business but has strict limitations on capital conversion.
• Companies are strictly prohibited from maintaining non-statutory accounting books or depositing corporate capital into individual accounts.

The company shall have no accounting books other than the statutory books. The company’s capital shall not be deposited in any account opened under the name of an individual.

• Companies must engage accounting firms qualified under PRC Securities Law for auditing and asset verification services.
• Shareholders or the board hold the power to appoint or dismiss accounting firms, but firms must be allowed to make representations during dismissal votes.
• The PRC Company Law strictly prohibits profit distribution until all prior losses are covered and statutory common reserve funds are allocated.
• Amendments to a company's articles of association require a supermajority vote of at least two-thirds of the shareholders present.
• Mandatory dissolution occurs under specific conditions, including the expiration of the company's term, shareholder resolution, or revocation of the business license.

The company should provide true and complete accounting evidence, accounting books, financial and accounting reports and other accounting information to the engaged accounting firm without any refusal, withholding or falsification of information.

• Shareholders holding at least 10% of voting rights can petition a court to dissolve a company if management deadlock causes significant losses.
• Dissolved companies must publicly announce their status on the National Enterprise Credit Information Publicity System within ten days.
• A company may avoid dissolution by amending its articles of association with a two-thirds majority vote, provided no assets have been distributed.
• Directors are legally obligated to form a liquidation committee within fifteen days of a dissolution event to manage assets and settle debts.
• The liquidation committee holds broad powers, including asset disposal, tax payment, and representing the company in civil legal proceedings.
• Creditors must be notified within ten days and have a specific window of up to forty-five days to lodge their claims with the committee.

If a liquidation committee is not established within the prescribed period or the liquidation fails to effect after the establishment of a liquidation committee, the interested party may file an application with a people’s court.

• Liquidation procedures mandate that company assets must satisfy taxes and debts before any distribution to shareholders.
• If assets are insufficient to cover liabilities during liquidation, the committee must petition a people's court for bankruptcy.
• Liquidation committee members are legally bound to act in good faith and are personally liable for losses caused by willful default or bribery.
• Overseas listings and subsequent offerings require strict filing with the CSRC within three business days of application or completion.
• Shareholders who lose registered certificates must obtain a court declaration of invalidity before the company can issue replacements.

Members of the liquidation committee shall be prohibited from abusing their authority in accepting bribes or other unlawful income and from misappropriating the company’s properties.

• Corporate mergers in the PRC occur via absorption or consolidation, requiring the dissolution of the absorbed or original entities.
• Companies must notify creditors within ten days of a merger resolution, allowing creditors to demand debt repayment or guarantees.
• Simplified merger procedures exist for 90% owned subsidiaries or small-scale acquisitions, bypassing shareholder meetings in favor of board approval.
• Post-division liabilities are jointly borne by the resulting companies unless a specific written agreement with creditors states otherwise.
• The China Securities Regulatory Commission (CSRC) serves as the primary regulatory body overseeing public offerings and market trading.
• The PRC Securities Law provides the first national framework for information disclosure and the acquisition of listed companies.

The liabilities of the company which have accrued prior to the division shall be jointly borne by the separated companies, unless otherwise stipulated in the agreement in writing.

• The PRC Securities Law governs all domestic and overseas issuance and trading of securities by Chinese enterprises.
• The PRC Arbitration Law establishes that arbitration agreements generally preclude parties from seeking resolution in people's courts.
• Arbitral awards are final and binding, with enforcement handled by people's courts unless specific legal violations are proven.
• Grounds for non-enforcement include forged evidence, lack of jurisdiction, or arbitrator misconduct such as bribery and malpractice.
• Enforcement of an award can be denied if a people's court determines it violates the public interest.
• Foreign arbitral awards may be recognized in China based on international treaties or the principle of reciprocity.

If the people’s court determines that the enforcement of the award violates the public interest, the award shall not be enforced.

• The PRC's accession to the New York Convention allows for the international recognition and enforcement of foreign arbitral awards based on reciprocity.
• China limits the application of the New York Convention specifically to disputes arising from mercantile legal relations and those involving other contracting states.
• A specialized mutual enforcement arrangement exists between Mainland China and Hong Kong, ensuring that arbitration awards from one jurisdiction are legally binding in the other.
• Recent supplemental arrangements allow parties to apply for enforcement in the court where a respondent is domiciled or where their property is located.
• A 2019 arrangement, effective in 2024, expanded the scope of reciprocal recognition for civil and commercial judgments between the Mainland and Hong Kong.
• The updated judicial framework removes the previous requirement for a specific 'choice of court' agreement, streamlining the enforcement process.

The arrangement discontinued the requirement for a choice of court agreement for bilateral recognition and enforcement.

• The Articles of Association outline the governance and share structure effective upon the Company's H share listing on the Hong Kong Stock Exchange.
• All shares are issued with a par value of RMB0.10 and must be issued under equitable conditions and pricing for the same class.
• Domestic and overseas listed shares (H shares) are granted identical rights regarding dividend distributions and other forms of capital return.
• The Company is permitted to increase capital through various methods, including bonus shares and capital reserve conversion, subject to shareholder approval.
• The board of directors may be authorized to issue up to fifty percent of existing shares within a three-year window without further shareholder meetings.
• Share repurchases are strictly prohibited except under specific conditions such as capital reduction, mergers, or employee incentive plans.

The Company shall issue shares in an open, fair and just manner, and each share of the same class issued at the same time shall be issued under the same conditions and at the same price per share.

• The Company is authorized to repurchase its own shares to safeguard shareholder value, facilitate employee stock plans, or convert corporate bonds.
• Repurchases must be conducted via open centralized trading or other methods approved by the CSRC and Hong Kong regulatory authorities.
• Specific repurchase scenarios require either a shareholder resolution or a two-thirds majority board approval depending on the underlying purpose.
• Repurchased shares must be cancelled or transferred within strict timelines ranging from ten days to three years based on the repurchase category.
• Treasury shares must be held in a segregated account and are stripped of voting rights, dividends, and other distribution entitlements.
• Any cancellation of shares necessitates a formal application to the company registry to reduce the registered capital by the nominal value of the shares.

The Company shall not exercise any rights attaching to treasury shares, including but not limited to voting rights, nor shall it declare or pay any dividends or other distributions in respect thereof.

• Pre-IPO shares are subject to a mandatory one-year lock-up period following the company's public listing.
• Directors and senior officers face strict annual transfer limits of 25% and post-employment cooling-off periods.
• The register of members serves as the definitive legal evidence of share ownership and associated rights.
• Shareholders possess fundamental rights to dividends, voting, and the supervision of business operations.
• Dissenting shareholders in merger or division scenarios have the specific right to demand a share buyback.
• Regulatory provisions from the CSRC and Hong Kong Stock Exchange take precedence over general transfer rules.

The aforesaid persons shall not transfer their shares in the Company within half a year after they terminate service with the Company.

• Shareholders are legally liable for compensation if they abuse their rights to cause losses to the company or other stakeholders.
• The corporate veil can be pierced, making shareholders jointly and severally liable for company debts if they use limited liability to evade repayment.
• Controlling shareholders and de facto controllers are bound by fiduciary and diligence obligations similar to those of directors.
• The shareholders' meeting serves as the ultimate authority, overseeing major decisions like capital changes, mergers, and profit distribution.
• Specific high-value transactions, such as asset sales exceeding 30% of total assets, require explicit approval from the shareholders' meeting.
• The shareholders' meeting must be held annually, with extraordinary meetings convened under specific circumstances within a two-month window.

Where shareholders of the Company abuse the Company’s position as an independent legal person and the limited liability of shareholders for the purposes of evading repayment of debts, thereby materially impairing the interests of the creditors of the Company, such shareholders shall be jointly and severally liable for the debts owed by the Company.

• The text outlines specific triggers for convening extraordinary shareholders' meetings, such as when the number of Directors falls below the legal minimum.
• A meeting must be called if the company's unrecovered losses reach one-third of its total paid-up share capital.
• Shareholders holding at least ten percent of voting shares have the legal right to request a meeting.
• The Board of Directors and the Audit Committee maintain the authority to initiate meetings when deemed necessary or appropriate.
• Independent directors can propose meetings, requiring a majority agreement among themselves to move forward.
• The Board must provide written feedback within ten days of a proposal and issue a notice within five days if the meeting is approved.

If the Board of Directors does not agree to hold an extraordinary shareholders’ meeting, it shall state reasons.

• The Audit Committee can formally propose extraordinary shareholders meetings to the Board of Directors in writing.
• The Board must respond within 10 days; failure to do so allows the Audit Committee to convene and preside over the meeting independently.
• Shareholders holding at least 10% of voting shares possess the legal right to request an extraordinary meeting from the Board.
• If both the Board and the Audit Committee fail to act on a valid request, 10% shareholders may convene the meeting themselves after a 90-day holding period.
• Any changes to the original meeting proposal require the specific consent of the initiating party, whether the Audit Committee or the shareholders.

In case the Board of Directors refuses to convene an extraordinary shareholders’ meeting, or does not give any response within 10 days upon receipt of the proposal, the Board of Directors shall be deemed to be unable or have failed to perform its duty.

• Proposals must fall within the scope of the shareholders' meeting and comply with all applicable laws and listing rules.
• The Board of Directors, Audit Committee, and shareholders holding at least 1% of shares have the right to submit proposals.
• Ad hoc proposals must be submitted in writing at least ten days before a meeting, triggering a supplementary notice within two days.
• Annual meetings require 20 days' prior written notice, while extraordinary meetings require 15 days' notice.
• Once a notice is issued, meetings cannot be postponed or cancelled without proper reasons and a two-day advance notification.
• The calculation of notice periods includes the date of issuance but excludes the actual date of the meeting.

Except as provided in the preceding paragraph, the convener of a shareholders’ meeting shall not amend the proposals set out in the notice of the shareholders’ meeting or put up any new proposals after the issuance of the notice of the shareholders’ meeting.

• Shareholders with weighted voting rights or their proxies may attend meetings and exercise voting rights according to company articles and listing rules.
• Specific protocols dictate who presides over meetings, ranging from the Board Chairman to Audit Committee members or shareholder-nominated representatives.
• If a presiding officer violates procedures and stalls a meeting, shareholders can vote to appoint a new presider to continue the session.
• Attendance requires strict identification protocols, including identity cards for individuals and legal authorization for corporate representatives.
• Resolutions are categorized as ordinary, requiring a simple majority, or special, requiring a two-thirds majority of the voting rights present.
• Ordinary resolutions cover essential business functions such as work reports, profit distribution, director remuneration, and auditor appointments.

When a shareholders’ meeting is convened, if the presider of the meeting contravenes the rules of procedure, rendering the meeting impossible to proceed, with the consent from more than half of the attending shareholders with voting rights, one person may be nominated at the shareholders’ meeting to serve as the presider and the meeting may proceed.

• Special resolutions are mandatory for fundamental corporate changes including capital adjustments, mergers, and company liquidations.
• Major asset disposals or guarantees exceeding 30% of total assets within a single year require high-level shareholder approval.
• Voting follows a one-vote-per-share principle, though exceptions exist for preferred shares with weighted voting rights.
• Minority investor interests are protected through separate vote counting and public disclosure for material matters.
• Connected shareholders are strictly prohibited from voting on transactions where they have a conflict of interest.
• Treasury shares held by the company itself are excluded from voting counts and do not carry any voting rights.

When the shareholders’ general meeting considers material matters affecting the interests of minority investors, the votes cast by such minority investors shall be counted separately.

• Directors are elected by shareholders for three-year renewable terms and can be removed prior to term completion by a shareholder vote.
• The Board of Directors consists of nine members, including one employee representative and three independent directors, with a cap on senior management representation.
• Incumbent directors must continue their duties beyond their term expiry if a successor has not yet been elected to ensure continuity of governance.
• The Board holds broad authority over corporate strategy, including profit distribution, capital adjustments, and the appointment of senior management.
• Specific committees for Audit, Nomination, Remuneration, and Strategy are mandated to support the Board's oversight functions.

The existing director shall continue to perform the duties of a director in accordance with laws, administrative regulations, departmental rules and the Articles of Association after the expiry of his/her term if no re-election is held in time.

• ESG Special Committees are established under the Board of Directors to perform duties according to specific implementation rules.
• The Secretary of the Board is designated as a senior management member responsible for meeting preparation, document custody, and information disclosure.
• A General Manager is appointed by the Board for three-year renewable terms to lead production operations and implement annual investment plans.
• The General Manager holds the power to devise internal management structures and propose the appointment or dismissal of other senior officers.
• Financial and accounting systems must strictly adhere to national laws and the specific listing rules of the stock exchange where the company is listed.
• The company is prohibited from maintaining non-statutory accounting books or storing corporate assets in individual accounts.

The assets of the company shall not be stored in any individual’s account.

• The company must allocate 10% of annual after-tax profits to a statutory reserve fund until it reaches 50% of registered capital.
• Current profits must be used to cover preceding year losses before any statutory reserve allocations are made.
• Remaining profits after reserves and loss coverage are distributed to shareholders based on their respective holdings.
• Dissolution can be triggered by term expiry, shareholder resolution, mergers, legal revocation, or significant operational hardship.
• Shareholders holding 10% of voting rights can petition the People's Court for dissolution if the company faces unresolvable losses.
• Upon dissolution, a liquidation group—typically composed of directors—must be formed within fifteen days to settle debts and dispose of assets.

If the Company suffers significant hardship in its operation and management, and the ongoing existence would bring significant losses for shareholders that cannot be resolved through other means, the shareholders holding at least ten percent of the total voting rights of the Company may request the People’s Court to dissolve the Company.

• The liquidation group is responsible for notifying creditors, managing civil litigation, and preparing a final distribution plan for shareholder or court approval.
• Company assets must be prioritized to pay liquidation expenses, employee wages, social insurance, taxes, and debts before any distribution to shareholders.
• During the liquidation process, the company maintains its legal existence but is strictly prohibited from engaging in business activities unrelated to its winding down.
• The Articles of Association must be amended if they conflict with updated laws, if company records change, or if the shareholders' meeting mandates a revision.
• The company transitioned from a limited liability entity to a joint stock company in 2025, with a subsequent share subdivision to adjust nominal share value.
• A principal place of business has been established in Hong Kong, subjecting the company to specific registration and service of process requirements under the Companies Ordinance.

During the liquidation period, the Company still exists but shall not carry out any business activities not related to liquidation.

• The document details the rapid formation and capital restructuring of various subsidiaries between 2024 and 2025.
• Several entities experienced massive capital injections shortly after their founding, such as the Tianjin branch increasing from RMB50 million to RMB950 million in two months.
• The geographic footprint of the company expanded across major Chinese hubs including Shanghai, Beijing, Zhuhai, Chengdu, and Zhejiang.
• The company established an international presence with new entities registered in the United Kingdom, Malaysia, and Europe.
• The diverse naming conventions suggest a focus on technology, education, culture, and media sectors.
• The rapid scaling of registered capital indicates a period of aggressive financial growth and preparation for large-scale operations.

On December 20, 2024, the registered capital of Tianjin Knowledge Atlas Technology Co., Ltd. increased from RMB50.0 million to RMB950.0 million.

• The establishment of two new subsidiaries, Shenzhen Knowledge Lingxin and Huangshi Knowledge Atlas, in late 2025 with varying capital scales.
• Shareholders approved a ten-for-one share subdivision to be executed immediately prior to the company's public listing.
• The Global Offering is structured to consist of no more than 15% of the total enlarged issued share capital.
• The Board received a general mandate to repurchase up to 10% of the total issued H Shares following the Listing.
• A separate mandate allows the Board to allot or issue up to 20% of the total shares in issue as of the Listing date.
• The company formally adopted new Articles of Association contingent upon the completion of the Global Offering.

the sub-division of the Shares with nominal value of RMB1.00 each on the basis of one to ten, effective immediately prior to the Listing

• The Board proposes a general mandate to repurchase up to 10% of the company's issued H Shares to bolster investor confidence and market reputation.
• Repurchased shares may be canceled or held as treasury shares depending on market conditions and capital management needs.
• The mandate is subject to shareholder approval via special resolution and expires at the conclusion of the next annual general meeting.
• Funding for the buybacks will be sourced from internal resources, including surplus funds and retained profits, in compliance with PRC laws.
• Strict blackout periods apply, prohibiting share repurchases for one month prior to the announcement of financial results or when inside information exists.

A listed company shall not repurchase its shares on the Stock Exchange at any time after inside information has come to its knowledge until the information is made publicly available.

• Directors and core connected persons have no current intention to sell H Shares back to the company if the repurchase mandate is approved.
• Repurchased H Shares will either be canceled or held as treasury shares in accordance with the Articles of Association and Listing Rules.
• A share repurchase that increases a shareholder's proportionate voting rights may trigger mandatory offer obligations under the Takeovers Code.
• The company will implement interim measures for treasury shares in CCASS to ensure it does not exercise voting rights or receive dividends.
• Directors undertake not to exercise the repurchase mandate to an extent that would materially harm the company's working capital or gearing position.
• The company currently holds no treasury shares and will not hold any upon its initial listing.

If, as a result of any repurchase of H Shares, a Shareholder’s proportionate interest in the voting rights of our Company increases, such increase will be treated as an acquisition for the purposes of the Takeovers Code.

• The text details multiple capital increase agreements executed on February 7, 2024, involving a wide array of technology companies and investment partnerships.
• Suzhou Junlian Jinfan Venture Capital and the Social Security Zhongguancun Innovation Investment Fund committed significant capital for newly increased registered shares.
• Shanghai Feiya Technology Co., Ltd. entered a major agreement to subscribe for registered capital at a consideration of RMB 200,000,000.
• The agreements involve a complex network of stakeholders including individual founders, specialized tech firms like Zhipu Overseas Innovation, and various limited partnerships.
• The participation of the AI Fund Partnership (Beijing) highlights a strategic focus on financing artificial intelligence and knowledge-based technology sectors.

Shanghai Feiya Technology Co., Ltd. (上海飛玡科技有限公司) agreed to subscribe for our Company’s newly increased registered capital of RMB465,090 at a consideration of RMB200,000,000.

• The document details multiple capital increase agreements involving the company and various technology-focused investment entities.
• A significant investment of US$30,000,000 was secured from Aramco Ventures Investments Limited in February 2024.
• The agreements involve a complex network of subsidiaries and partners, including Nanjing Zhihu and various 'Knowledge' (Zhipu) branded technology firms.
• Individual stakeholders such as Tang Jie and Li Juanzi are consistently named as parties to these multi-lateral corporate restructuring and funding contracts.
• Funding rounds occurred in rapid succession throughout 2024, specifically in February, April, and August, indicating aggressive capital expansion.

Aramco Ventures Investments Limited, pursuant to which Aramco Ventures Investments Limited agreed to subscribe for our Company’s newly increased registered capital of RMB495,390 at a consideration of US$30,000,000.

• The text details multiple high-value capital increase agreements involving various technology and investment partnerships.
• Beijing Zhongguancun Science City Phase II agreed to a consideration of RMB500,000,000 for newly increased registered capital.
• Beijing Daxing Industrial Fund Partnership committed RMB300,000,000 to subscribe for company capital in August 2024.
• The agreements involve a complex network of subsidiaries and individual stakeholders including Tang Jie and Li Juanzi.
• Beijing Xinglian Zhaoji Enterprise Management Partnership entered a Series B6-1 agreement for a consideration of RMB150,000,000.

Beijing Zhongguancun Science City Phase II Technology Growth Equity Investment Partnership (Limited Partnership) agreed to subscribe for our Company’s newly increased registered capital of RMB826,211 at a consideration of RMB500,000,000.

• The document details a series of complex capital increase agreements involving various technology entities and investment partnerships.
• Major financial injections include a significant RMB 950,000,000 subscription by the Tianjin Haihe Fuxin Youda Venture Capital Fund.
• A convertible note agreement was established in September 2024, providing a principal loan amount of RMB 130,000,000 to the company.
• The agreements involve a wide network of 'Zhipu' branded subsidiaries across Beijing, Shanghai, Shenzhen, and overseas locations.
• Key individual stakeholders and academic-linked figures such as Tang Jie and Li Juanzi are explicitly named as parties to these legal instruments.
• Institutional backing is evidenced by the participation of the Social Security Zhongguancun Innovation Investment Fund.

Tianjin Haihe Fuxin Youda Venture Capital Fund Partnership (Limited Partnership) agreed to subscribe for our Company’s newly increased registered capital of RMB1,569,800 at a consideration of RMB950,000,000.

• The document details complex capital increase agreements involving numerous Zhipu-affiliated technology entities across various Chinese cities.
• A significant capital reduction agreement was reached with Beijing Xinglian Zhaoji, adjusting registered capital to RMB160,285 for a consideration of RMB97 million.
• Chengdu High-tech Orrino Youchan Equity Investment Fund committed a substantial investment of RMB300 million for newly increased registered capital.
• The text identifies key individual stakeholders and academic-linked founders including Tang Jie, Li Juanzi, and Liu Debing.
• Multiple convertible note agreements were executed in early 2025, securing additional funding of at least RMB100 million from specialized investment partnerships.

Chengdu High-tech Orrino Youchan Equity Investment Fund Partnership (Limited Partnership) agreed to subscribe for our Company’s newly increased registered capital of RMB495,726 at a consideration of RMB300,000,000.

• The document details a series of significant capital increase agreements involving the Company and numerous technology-focused subsidiaries.
• A major agreement dated January 27, 2025, involves a capital injection of RMB350,000,000 from Hangzhou Chengtou Industrial Development Investment.
• Hangzhou Shangcheng Linghang Venture Capital also participated in the January round with a consideration of RMB100,000,000.
• The agreements involve a complex network of stakeholders, including individual founders and various limited partnerships across major Chinese tech hubs.
• A subsequent agreement on February 7, 2025, introduces institutional investors like the Social Security Zhongguancun Innovation Investment Fund.

Hangzhou Chengtou Industrial Development Investment Partnership (Limited Partnership) agreed to subscribe for our Company’s newly increased registered capital of RMB578,347 at a consideration of RMB350,000,000.

• The text details several high-value capital increase agreements involving major investment funds and the Company's registered capital.
• Significant convertible note agreements were established in early 2025, including a RMB100 million loan from the Lenovo Small and Medium Enterprise Development Venture Capital Fund.
• A substantial convertible loan of RMB400 million was provided by the Zhuhai Huafa New Quality Productivity Investment Fund in March 2025.
• The AI Fund Partnership (Limited Partnership) committed to a capital subscription of RMB200 million, involving a complex network of technology subsidiaries.
• The agreements involve a wide array of regional technology entities across China, including branches in Beijing, Shanghai, Shenzhen, and Hangzhou.
• Key individuals, including Tang Jie and Liu Debing, are named as central parties to these multi-million RMB financial instruments and investment contracts.

Zhuhai Huafa New Quality Productivity Investment Fund Partnership (Limited Partnership) agreed to provide a convertible loan to the Company in the principal amount of RMB400,000,000.

• The document details complex capital increase agreements involving a vast network of subsidiary technology companies across major Chinese cities.
• A significant investment of RMB100,000,000 was made by the Lenovo Small and Medium Enterprise Development Venture Capital Fund.
• A second major capital injection of RMB130,000,000 was executed by the Tianjin Tianchuang Haihe Yongtai Puxin Venture Capital Partnership.
• Both major investments were satisfied through the conversion of existing convertible notes into newly increased registered capital.
• The agreements involve a diverse group of stakeholders including individual founders, limited partnerships, and international entities in Hong Kong and Singapore.

Lenovo Small and Medium Enterprise Development Venture Capital Fund (Tianjin) Partnership (Limited Partnership) agreed to subscribe for our Company’s newly increased registered capital of RMB165,242 at a consideration of RMB100,000,000, which was satisfied through the conversion of convertible note.

• The text details complex legal agreements involving a vast network of 'Knowledge Atlas' (Zhipu Huazhang) subsidiaries across major Chinese cities.
• A significant capital increase was executed through the conversion of a convertible note by Hainan Xiarui Investment Partnership.
• The agreements involve high-level individual stakeholders and numerous limited partnership investment vehicles.
• A second major capital increase agreement dated May 13, 2025, involves the Zhuhai Huafa New Quality Productivity Investment Fund.
• The geographical spread of the entities—from Beijing and Shanghai to Hong Kong and Singapore—highlights a broad corporate structure.

Hainan Xiarui Investment Partnership (Limited Partnership) (海南夏睿投資合夥企業(有限合夥)) agreed to subscribe for our Company’s newly increased registered capital of RMB165,242 at a consideration of RMB100,000,000, which was satisfied through the conversion of convertible note;

• The document details a significant capital increase agreement dated May 23, 2025, involving a wide network of technology subsidiaries.
• A convertible note was satisfied through the conversion of RMB500,000,000 into company equity.
• Zhihui Linghang Venture Capital Partnership agreed to subscribe for newly registered capital at a substantial premium, paying RMB500,000,000 for RMB826,211 in registered capital.
• The agreements involve a complex web of entities across major Chinese tech hubs including Beijing, Shanghai, Shenzhen, and Hangzhou.
• Key individual stakeholders and academic-linked figures, such as Tang Jie and Li Juanzi, are listed as parties to the comprehensive shareholders' agreement.

Zhihui Linghang Venture Capital Partnership (Limited Partnership) (上海浦東智慧領航創業投資合夥企業(有限合夥)) agreed to subscribe for our Company’s newly registered capital of RMB826,211 at a consideration of RMB500,000,000.

• The text provides an extensive list of corporate entities and limited partnerships involved in a significant financial or legal transaction.
• A diverse range of investment vehicles is represented, including private equity firms, venture capital centers, and industrial investment funds.
• Major Chinese technology and education giants are listed as stakeholders, including subsidiaries of Tencent, Meituan (Sankuai), and TAL Education.
• The geographic distribution of these entities spans major Chinese economic hubs such as Beijing, Shenzhen, Shanghai, Hangzhou, and Tianjin.
• The inclusion of the Social Security Zhongguancun Innovation Investment Fund indicates participation from state-linked or institutional social security capital.

Social Security Zhongguancun Innovation Investment Fund (Beijing) Partnership (Limited Partnership) (社保基金中關村自主創新投資基金(北京)合夥企業(有限合夥))

• The document details a complex network of investment partnerships and technology companies involved in a shareholder special rights termination agreement.
• Key participants include major venture capital entities such as Lenovo Small and Medium Enterprise Development and various Qiming and Junlian investment arms.
• The agreement, dated June 27, 2025, involves a significant consolidation or restructuring of rights across numerous 'Knowledge Atlas' (Zhipu) technology subsidiaries.
• The list of parties includes both corporate entities and individual stakeholders like Tang Jie and Li Junazi, indicating a broad consensus for the legal transition.
• The geographical scope of the involved entities spans major Chinese tech hubs including Beijing, Shenzhen, Shanghai, and Hangzhou, as well as international offices in Hong Kong and Singapore.
• This legal framework is a critical step in the company's preparation for a change in corporate status, likely preceding an IPO or major acquisition.

(aa) a shareholder special rights termination agreement dated June 27, 2025 entered into among our Company, Nanjing Knowledge Atlas Technology Co., Ltd. (南京智譜華章科技有限公司), Shenzhen Knowledge Atlas Technology Co., Ltd. (深圳智譜華章科技有限公司)...

• The document lists an extensive network of venture capital and equity investment partnerships involved in a corporate restructuring or IPO process.
• Major institutional players identified include Tencent Venture Capital, TAL Education, and various regional industrial development funds from Beijing, Hangzhou, and Shenzhen.
• A significant legal action involves the formal termination of certain shareholder rights among a large group of limited partnerships and technology firms.
• Cornerstone investment agreements have been established with major entities like JSC International Investment Fund and JinYi Capital.
• The financial commitments from cornerstone investors involve substantial sums, including a subscription of H Shares valued at approximately US$179 million.
• The involvement of the Social Security Zhongguancun Innovation Investment Fund indicates high-level state-backed institutional support.

pursuant to which the parties agreed to terminate certain shareholder rights

• The text details a series of cornerstone investment agreements finalized in late December 2025 for an H Share offering.
• Major financial entities including WT Asset Management, Taikang Life Insurance, and GF Management are committing significant capital ranging from US$3.55 million to US$44.9 million.
• China International Capital Corporation (CICC) acts as a central intermediary and securities dealer across nearly all listed agreements.
• Specific financial structures are mentioned, such as CICC Financial Trading Limited subscribing to shares to hedge cross-border delta-one OTC swap transactions.
• The document transitions from listing these material contracts into a section regarding the company's registered intellectual property rights.

CICC Financial Trading Limited has agreed to subscribe for H Shares at the Offer Price in the aggregate amount of Hong Kong dollar equivalent of US$9.0 million and hold such H Shares on a non-discretionary basis to hedge a series of cross border delta-one OTC swap transactions.

• The document lists 22 registered trademarks held by the company in the PRC and Hong Kong, with most PRC registrations expiring in 2034.
• The company holds a significant portfolio of invention patents specifically focused on Large Language Models (LLMs) and AI agent training.
• Key patented technologies include methods for optimizing instruction-following capabilities and dynamically adjusting the depth of LLMs.
• The portfolio covers specialized AI infrastructure such as Mixture of Experts (MoE) models based on decision trees and automatic parallelized text generation.
• Several patents focus on evaluation and alignment, including specific methods for assessing Chinese LLMs and multi-dialogue systems.
• The intellectual property is categorized as material to the company's business operations and statutory standing.

As of the Latest Practicable Date, we had registered the following patents which we consider to be or may be material to our business.

• The document lists several invention patents held by Beijing Lingxin Intelligent Technology, focusing on human-machine dialogue and personality testing.
• A significant portion of the intellectual property involves Large Language Models (LLMs), including safety evaluation methods and model optimization systems.
• The company has registered multiple software copyrights for its core products, such as AutoGLM, ChatGLM-6B, and the GLM3.5_130B platform.
• The portfolio includes specialized technology for extracting key information frames from surveillance videos via Nanjing Knowledge Atlas Technology.
• Strategic domain names like bigmodel.cn and chatglm.cn have been secured with long-term registrations extending into the 2030s.

A personality test method, device and electronic equipment based on human-machine dialogue (基於人機對話的性格測試方法、裝置及電子設備)

• The document outlines the disclosure of interests and short positions held by directors, supervisors, and chief executives following the Global Offering.
• Regulatory compliance is maintained through adherence to the Securities and Futures Ordinance (SFO) and the Model Code for Securities Transactions.
• Substantial shareholders holding 10% or more of any class of voting share capital must be disclosed under specific listing rules.
• Beijing Knowledge Huixing Technology Co., Ltd. is identified as a significant shareholder with a 30% stake held by Hainan Hezun Investment Co., Ltd.
• Directors have entered into three-year service contracts that align with the term of the Board and are subject to renewal and shareholder approval.

Save as disclosed above, we have not entered, and do not propose to enter, into any service contracts with any of our Directors or Supervisor in their respective capacities as Directors.

• The document confirms that directors and supervisors received no undisclosed remunerations or benefits in kind during the specified track record period.
• Strict disclaimers state that no directors or experts hold material interests in recent asset acquisitions, leases, or significant business contracts.
• Directors and major shareholders (owning over 5%) are confirmed to have no interests in the company's top five customers or suppliers.
• The company has established three Employee Incentive Schemes (2021, 2023, and 2025) utilizing Employee Ownership Platforms.
• Employee Ownership Platforms hold approximately 16.55% of issued shares, which will adjust to 15.15% following the Global Offering.
• The incentive schemes are structured to avoid further shareholder dilution after listing, as all relevant shares have already been issued.

The Employee Incentive Schemes will not cause any dilution of the shareholding of our Shareholders after the Listing given all Shares under the Employee Incentive Schemes have been issued to the Employee Ownership Platforms.

• The Employee Incentive Schemes aim to align the interests of key personnel with the Group's long-term growth and governance performance.
• Eligible participants include senior and middle management, core technical staff, and essential consultants designated as 'Grantees'.
• The CEO Office administers the schemes, determining participant eligibility and the specific terms of Restricted Awards.
• A mandatory lock-up period prevents the transfer of awards until at least 12 months after the company's official listing.
• Exiting employees may retain vested awards, but unvested portions are repurchased at the original subscription price.
• Statutory disclosures confirm the absence of material litigation and the independence of the Sole Sponsor for the Global Offering.

Each of the Grantees may not transfer the Restricted Awards, and the Employee Ownership Platforms may not transfer, pledge or otherwise dispose of the relevant underlying Shares.

• The document lists legal advisors and industry consultants who have provided written consent for their reports and opinions to be included in the prospectus.
• None of the named experts hold shareholding interests or rights to subscribe for securities in the Group as of the latest practicable date.
• A Compliance Advisor, Maxa Capital Limited, has been appointed to ensure adherence to Listing Rule 3A.19.
• The text details Hong Kong stamp duty obligations for H Shares, including a 0.10% ad valorem rate for both buyers and sellers and potential penalties for non-payment.
• The company confirms there has been no material adverse change in its financial or trading position since the last audited statements on June 30, 2025.
• The prospectus establishes a binding effect under the Companies Ordinance for all applicants, except for specific penal provisions.

If no stamp duty is paid on or before the due date, a penalty of up to ten times the duty payable may be imposed.

• The text outlines specific financial declarations, confirming no commissions or special terms have been granted for share sales beyond what is disclosed in the underwriting section.
• It certifies the operational stability of the Group, stating there have been no significant business interruptions in the twelve months preceding the prospectus.
• The document confirms that the company currently holds no outstanding convertible debt securities, debentures, or deferred management shares.
• Arrangements have been finalized for H Shares to be admitted into the Central Clearing and Settlement System (CCASS) for efficient clearing and settlement.
• A comprehensive list of documents, including the Articles of Association and audited financial statements, is made available for public inspection on official websites.
• Legal opinions and memorandums from various jurisdictions, including the PRC and Singapore, are provided to validate corporate matters and data compliance.

There has not been any interruption in the business of our Group which may have or has had a significant effect on the financial position of our Group in the 12 months preceding the date of this prospectus.

• The document lists formal legal memorandums provided by King & Wood Mallesons regarding international sanctions.
• It references an industry-specific report compiled by Frost & Sullivan to provide market context.
• The text includes written consents from various experts as detailed in the statutory and general information appendix.
• Key Chinese regulatory frameworks are cited, including the PRC Company Law and the PRC Securities Law.
• The section highlights the administrative measures governing overseas securities offerings by domestic Chinese companies.
• Unofficial English translations of these primary Chinese laws are provided for reference.

the Trial Administrative Measures of Overseas Securities Offering and Listing by Domestic Companies, together with their unofficial English translation.

• The company is seeking a listing under Chapter 18C of the Listing Rules, a specific pathway for technology companies that do not meet standard financial requirements.
• Founded in 2019, the firm is a leading Chinese AI developer focused on Artificial General Intelligence (AGI) and the Model-as-a-Service (MaaS) platform.

In particular, we are a specialist technology company seeking to list on the Main Board of the Hong Kong Stock Exchange under Chapter 18C of the Listing Rules because we are unable to meet the requirements under Rule 8.05(1), (2) or (3) of the Listing Rules.

• The company ranks as China's largest independent general-purpose large model developer by revenue as of 2025.
• Their business model centers on a Model-as-a-Service (MaaS) platform offering language, multimodal, agent, and coding models.

Through this product development and commercialization platform, we deliver intelligence to institutional customers, developers and individual customers in the most suitable, sensible and scalable way despite great heterogeneity in computing infrastructure, devices and applications.

• GLM-4.5 is a 355-billion-parameter model that excels in agentic, reasoning, and coding tasks while supporting multi-modal extensions.
• The model ranked first among global open-source models and first in China across twelve industry-standard benchmark tests in July 2025.

Reflection and rumination models spend additional time “deep thinking” before generating an answer, which makes them better for complex reasoning tasks.

• AutoGLM marks a shift from conversational AI to 'acting' AI, capable of autonomously controlling digital devices via graphical user interfaces.
• AutoGLM 2.0 operates on the cloud, allowing the AI to complete tasks across mobile apps and websites without interrupting the user's local device usage.

AutoGLM represents a major step forward in the evolution of our AI universe—from “chat” to “act,” bridging the gap between conversation-based AI and real-world task execution.

• Revenue has demonstrated explosive growth with a CAGR exceeding 130%, rising from RMB57.4 million in 2022 to RMB312.4 million in 2024.
• Significant net losses have occurred alongside revenue growth, primarily driven by massive investments in research and development which reached RMB2,195.4 million in 2024.

Our losses during the Track Record Period were primarily due to our significant investments in research and development.

With our deep academic roots as cornerstone for technological leadership, we are constantly making technological advancements and iterations of our foundation models, boosting our profitability.

• The company is applying for a Hong Kong Stock Exchange listing under Rule 18C.03, targeting a market capitalization exceeding HK$6 billion.
• The business faces significant R&D risks, specifically the necessity of continuous innovation in the rapidly changing AI industry to remain competitive.

The development of AGI is still at an early stage and there are substantial uncertainties in the future realization of AGI.

Research and development expenses . . . . . . . . . . (84,377) (147.0) (528,884) (424.7) (2,195,436) (702.7)

• The company experienced a dramatic increase in net losses, rising from RMB143.7 million in 2022 to over RMB2.9 billion by 2024.
• Significant capital is being deployed into research and development to build advanced foundation models and expand the R&D team.

Such overall upward trend in our net loss was primarily due to (i) the significant increase in research and development expenses as we expanded our R&D team and procured substantially more computing services from third parties and related computing hardware.

• Net current liabilities increased significantly to RMB7,088.9 million by mid-2025, largely due to financial instruments issued to Pre-IPO investors.
• The monthly cash burn rate has escalated dramatically from RMB3.0 million in 2022 to RMB327.3 million in 2025, driven by R&D and hardware procurement.

Our historical cash burn rate was RMB3.0 million, RMB105.9 million, RMB194.5 million and RMB327.3 million for each of the years ended December 31, 2022, 2023 and 2024 and for the six months ended June 30, 2025, respectively.

• The company estimates a financial runway of 27.3 to 38.9 months depending on the utilization of net proceeds from its upcoming listing.
• The global offering is priced at HK$116.20 per share, targeting a total market capitalization of approximately HK$51.155 billion.

We will continue to monitor our cash flows from operations closely and maintain our financial viability through a variety of means, including, among others, banking facilities and external financings.

• The company expects to receive net proceeds of approximately HK$4,173.4 million based on an offer price of HK$116.20 per share.
• A significant majority of the proceeds, approximately 70%, is earmarked for strengthening research and development in general-purpose large AI models.

Investors should not purchase our ordinary shares with the expectation of receiving cash dividends.

• The company introduced AutoGLM, an autonomous agent powered by GLM-4.5 that simulates human actions on mobile and web applications without occupying the user's device.
• Institutional customer growth surged to over 12,000 by September 2025, with daily token consumption reaching 4.2 trillion in November 2025.

It can autonomously complete requested tasks on the cloud without occupying the user’s mobile phone or computer, allowing users to continue using their own devices without interruption.

• Investors are warned that the market price of H Shares could decline significantly, potentially leading to a total loss of investment.
• Material risks are categorized into seven key areas including R&D, commercialization, intellectual property, and jurisdictional challenges.

In any such case, the market price of our H Shares could decline, and you may lose all or part of your investment.

• The company's future success depends on its ability to integrate multimodal advancements and AI agents into its large language models.
• Rapid technological shifts or the emergence of alternative technologies could render current models obsolete or unattractive to the market.

In addition, the emergence of new or alternative technologies could replace or reduce the demand for our models and solutions, render our technologies obsolete or unattractive.

• Technological advancement is categorized into five stages of Large Language Models, with the company currently attempting to move beyond the first three.
• The realization of Artificial General Intelligence (AGI) remains speculative, and failure to achieve it could invalidate the company's massive capital investments.

However, we cannot guarantee that we will be able to the self-perception stage and consciousness stage.

• The LLM market is characterized by intense and increasing competition that is expected to lead to rapid industry consolidation.
• Pricing pressures resulting from market competition could lead to significant reductions in profitability and market share.

In light of these factors, even if our models are more advanced, effective and cost-efficient than those of our competitors, current or potential customers may accept competitors’ models in lieu of ours.

• The company's success depends on maintaining model compatibility across diverse hardware and software platforms, currently supporting over 40 global chip platforms.
• Compliance requirements include mandatory security assessments and algorithm filings for services with public opinion or social mobilization capabilities.

The interpretation and implementation of existing measures are evolving and the PRC regulatory agencies, including the CAC, may further adopt new laws, regulations, rules, or detailed implementation and interpretation related to the above-mentioned measures, which may negatively affect us.

• The PRC government has rapidly expanded its regulatory framework for cybersecurity and data protection through several major legislative acts since 2017.
• The Personal Information Protection Law (PIPL) establishes strict rules for the entire lifecycle of personal data processing.

There is no guarantee as to the effectiveness of the measures we have taken to urge or supervise such third-party vendors or public data controllers to abide by applicable cybersecurity, data protection and personal information protection laws and regulations.

• The company faces potential classification as a Critical Information Infrastructure Operator (CIIO), which would trigger mandatory national security reviews.
• New Chinese regulations mandate security assessments and certifications for cross-border transfers of personal information and important data.

Any such review process could result in increased compliance costs, delays to our business operations or product deployment, and may have a material adverse effect on our business, financial condition and results of operations.

• The company and nine subsidiaries were added to the U.S. Department of Commerce's Entity List in early 2025.
• Being on the Entity List restricts access to U.S.-origin goods, software, and technology, including certain non-U.S. items with U.S. content.

The bill received unanimous bipartisan support in the U.S. House Foreign Affairs Committee.

• New U.S. regulations, including Executive Order 14105 and the America First Investment Policy, restrict investments in Chinese tech firms.
• The company is classified as a 'covered foreign person,' leading to a total prohibition on U.S. persons participating in the Global Offering.

Accordingly, as advised by our International Sanctions Counsel, U.S persons (as defined under the Final Rule) are prohibited to invest in our Company in the Global Offering.

• Compliance with international laws is difficult, as meeting the legal requirements of one country may inadvertently violate the regulations of another.
• Intellectual property is critical to the business, yet existing legal protections may not be sufficient to prevent misappropriation by competitors or former employees.

In some cases, compliance with the laws and regulations of one country could violate the laws and regulations of another country.

• The company experienced rapid revenue growth between 2022 and 2025, driven primarily by the commercialization of general-purpose large AI models.
• Heavy investments in research and development for foundation models and aggressive sales marketing are the primary drivers of the accumulated losses.

We may not be able to sustain our historical growth rates, and our historical growth may not be indicative of our future growth or financial results.

• The Overseas Listing Trial Measures, effective March 31, 2023, mandate filing requirements for Chinese companies seeking to list securities abroad.
• Failure to comply with filing procedures or providing false information can result in administrative penalties, including warnings and significant fines.

If a domestic company fails to complete the filing procedure or conceals any material fact or falsifies any major content in its filing documents, such domestic company may be subject to administrative penalties.

• Strict PRC regulations on Renminbi conversion and remittance could limit the company's ability to pay dividends or fund offshore capital expenditures.
• Non-resident investors may be subject to a 10% mainland China withholding tax on dividends and capital gains from share transfers.

As a result, you may encounter difficulty in enforcing foreign judgments against us or our Directors or senior management.

• The Listing Rules permit existing shareholders of Specialist Technology Companies to participate in IPOs to meet funding needs.
• Strict anti-preference rules ensure that existing shareholders receive no special allocation advantages over other investors, aside from assured entitlement at the IPO price.

In the case of subscription as a placee, the applicant and its sponsors must confirm that no preference in allocation was given to the existing shareholder.

• The Company and Sole Sponsor confirm that no preferential treatment was given to specific cornerstone investors like Qizhi SP and Luster LightTech beyond assured entitlement at the Offer Price.
• The Stock Exchange granted a waiver for CICC FT's participation provided the terms are no more favorable than those offered to other cornerstone investors.

CICC FT will hold the Offer Shares on a non-discretionary basis to hedge the Gaoyi OTC Swaps, while the economic risks and returns of the underlying Offer Shares are passed to the CICC FT Ultimate Clients (Gaoyi).

• Trading is expected to commence at 9:00 a.m. on January 8, 2026, under the stock code 2513, with shares traded in board lots of 100.
• The H Shares will be eligible for deposit, clearance, and settlement through the Central Clearing and Settlement System (CCASS).

Under section 44B(1) of the Companies (Winding Up and Miscellaneous Provisions) Ordinance, any allotment made in respect of any application will be invalid if the listing of, and permission to deal in, the H Shares on the Stock Exchange is refused.

• Artificial Intelligence serves as the primary driver of the Fourth Industrial Revolution, enabling intelligent automation and the merging of physical and digital realms.
• China's AI market is experiencing rapid acceleration, with a projected compound annual growth rate of 35.5% leading into 2030.

Unlike the first three industrial revolutions which were powered by steam, electricity and computers and Internet, the Fourth Industrial Revolution is unfolding at a faster pace, on a broader scale, and with deeper impacts on society and the economy.

• The commercialization of Large Language Models (LLMs) in China is primarily driven by institutional and enterprise demand rather than individual consumers.
• The Chinese LLM market is projected to grow from RMB5.3 billion in 2024 to RMB101.1 billion by 2030, maintaining a robust CAGR of 63.5%.

In contrast, the enterprise scenario is the main driver of growth in LLM market in China.

• A positive feedback loop is created by using real-world client data to iteratively lower training costs while improving model performance.
• The creation of a 'closed-loop' ecosystem involving chip manufacturers, developers, and industry-specific clients forms a significant barrier to entry.

This forms a positive feedback loop featuring low training costs, strong model performance and high adaptability to applications.

• The Deep Synthesis Provisions mandate that service providers verify user identities and label AI-generated content to prevent the spread of illegal information.
• The Interim Measures on Generative AI hold service providers legally responsible as the 'producers' of any online information generated by their systems.

Service providers shall be responsible as producers of the online information.

• New regulations mandate that all AI-generated content must feature both explicit visual indicators and implicit digital watermarks.
• The company has completed all necessary algorithm and generative AI filings required by PRC legal authorities.

Explicit markings, such as visible text prompts and visual indicators, must be presented prominently to users.

• The Cybersecurity Review Measures mandate national security reviews for critical information infrastructure operators (CIIOs) and network platform operators.
• Network Platform Operators with over one million users must undergo a cybersecurity review before seeking a listing in a foreign country.

In addition, Network Platform Operators holding personal information of more than one million users that seek listing in a foreign country are obliged to apply for a cybersecurity review by the Cybersecurity Review Office.

• China employs a 'Negative List' system where foreign investment is prohibited or restricted in specific industries but permitted elsewhere under national treatment.
• A security review mechanism is mandatory for investments in national defense or major sectors where foreign investors acquire actual control.

The following categories of foreign investment shall require foreign investors or domestic relevant parties to proactively declare the security review prior to implementation: (i) investments in national defense security-related sectors... and (ii) investments in major sectors affecting national security.

• The Bureau of Industry and Security (BIS) added the company and nine subsidiaries to the Entity List on January 16, 2025, restricting access to items subject to the EAR.
• The Group has implemented comprehensive export control compliance measures across all entities to mitigate risks following the Entity List Addition.

The EAR defines “knowledge” as including “positive knowledge that the circumstance exists or is substantially certain to occur,” as well as “an awareness of a high probability of its existence or future occurrence,” which is “inferred from evidence of the conscious disregard of facts known to a person and is also inferred from a person’s willful avoidance of facts.”

• Founded in 2019 by a team from Tsinghua University, the company is a leading Chinese AI firm focused on achieving artificial general intelligence (AGI).
• The 2024-2025 roadmap shows a rapid expansion into multimodal capabilities, including visual comprehension, video generation, and emotional voice models.

Our Company draws on the academic rigor and innovative spirit of its founding team to advance cutting-edge developments in AI.

• The company underwent multiple funding rounds from Series Angel through Series B6, showing a rapid escalation in share price from RMB3.87 to RMB60.52.
• Pre-IPO investors received significant discounts to the offer price, ranging from 96.33% for early angel investors to 42.58% for the final Series B6 round.

The Pre-IPO Investors’ investment demonstrated their confidence in our Group and served as an endorsement of our performance, strengths and prospects.

• A formal Concert Party Agreement establishes a unified voting bloc among key founders and entities, ensuring collective control over the company.
• The group of Controlling Shareholders collectively holds approximately 33.03% of the company's shares prior to the Global Offering.

In particular, Dr. Liu and Dr. Tang will exercise their shareholder rights as Shareholders in relation to major matters of the Group consistently and collectively, and in the event of a disagreement or dispute, the final decision shall rest with Dr. Liu.

• The company utilizes a 'virtuous insight flywheel' by deploying models to understand real-world usage and refine training strategies.
• Open-source engagement is a primary driver of innovation, with over 45 million model downloads and 1,000 projects created by global developers.

We see open-sourcing as a key catalyst for the vitality of our ecosystem.

• The company is a pioneer in China's AI sector, being the first to develop large models with over 100 billion parameters and ranking first among independent Chinese developers by 2024 revenue.
• As of July 2025, their flagship GLM-4.5 model ranked third globally and first in China across twelve industry-standard benchmark tests.

In March 2025, we introduced AutoGLM-Rumination, a major breakthrough for frontier AI agents as it is able to autonomously carry out agent operations that require sustained “thinking while working” for tens of minutes.

• The company maintains a research-centric culture with over 74% of its workforce dedicated to R&D and a portfolio of 500 papers with 58,000 citations.
• The 'Model-as-a-Service' (MaaS) platform offers an all-in-one ecosystem including language, multimodal, and coding models like GLM-4.5 and CodeGeeX.

This association, rooted in a shared academic lineage, helps instill continuity and coherence in our technological roadmap.

• The Model-as-a-Service (MaaS) platform streamlines business workflows by offering customizable agent templates and scenario-based solutions.
• Significant cost efficiency is achieved through high-performance smaller models, such as GLM-4-Air, which reduces computing costs by 20 times compared to larger competitors.

For example, our reasoning model GLM-4-Air, with a model size of only 32 billion parameters, delivers a performance similar to competitors’ models with a size of 671 billion parameters, reducing computing costs by 20 times.

• The GLM-4.5 flagship foundation model serves as the core for a diverse ecosystem of specialized AI, including reflection, multimodal, and coding models.
• AI agents such as AutoGLM and CoCo integrate reasoning and tool-use to perform multi-step tasks autonomously without human intervention.

Reflection and rumination models spend additional time “deep thinking” before generating an answer, which makes them better for complex reasoning tasks.

• The GLM-4.5V model integrates language and vision to perform complex reasoning tasks beyond simple image recognition.
• The model can autonomously reproduce functional webpages, including static layouts and dynamic interactions, from screenshots or screen recordings.

GLM-4.5V excels as a “visual detective,” combining perceptual acuity with reasoning skills for image recognition and inference.

• AutoGLM 2.0 significantly outperforms leading models like GPT-4o and Claude-3.5-Sonnet in mobile and browser benchmarks.
• The system utilizes a cloud-based execution model that allows the agent to complete tasks autonomously without occupying the user's physical device.

Building capable foundation agents requires enriching them with dynamic knowledge, either through direct interaction with real-world environments or through learning from synthesized trajectories.

• CodeGeeX serves as a versatile AI coding assistant capable of generating 100 million lines of code daily as of mid-2025.
• Institutional customer growth has scaled rapidly, increasing from 48 customers in 2022 to over 3,100 by June 2025.

As of June 30, 2025, CodeGeeX generated on average more than 100 million lines of code on a daily basis.