Surging Hong Kong Stock AI Large Model Sector, Zhipu Rises 114% in a Month! | In-depth Analysis | Research Report | Potential Stocks | Hong Kong Stocks | AI

Since February 2026, China’s large AI model industry has seen a series of landmark developments, with industry momentum and growth potential continuously rising. The Hong Kong-listed large model sector has performed particularly strongly, becoming a core focus in the capital market. $KNOWLEDGE ATLAS (02513.HK)$ 、 $MINIMAX-WP (00100.HK)$Stock prices of core companies such as Zhipu (2513.HK) have repeatedly hit new highs. Specifically, due to the technological breakthrough of its new flagship model GLM-5, Zhipu’s stock price surged for two consecutive days, bringing its total market capitalization to HKD 210 billion. $ILUVATAR COREX (09903.HK)$ 、 $BIREN TECH (06082.HK)$Stock prices of compute-related enterprises have also risen in tandem, reflecting the capital market’s strong recognition of the long-term value of the industry. Meanwhile, leading technology firms have rolled out a wave of new-generation large model products, showcasing concentrated technical breakthroughs across several critical fields such as long-context processing, programming optimization, and multimodal creation. Zhipu's (2513.HK) GLM-5 ranked first globally in single-test performance for agent-based programming, surpassing Claude Opus. MiniMax-WP (0100.HK) launched M2.5, the world's first production-grade programming model natively designed for Agent scenarios, while other notable products like Qwen 3.5 and ByteDance’s SeeDance 2.0 were unveiled, accelerating the pace of technological iteration.

圖一： $KNOWLEDGE ATLAS (02513.HK)$Price Increase Since Listing

Recently, the Hong Kong-listed large model sector has shown a strong upward trend, with the 'first domestically produced large model stock' $KNOWLEDGE ATLAS (02513.HK)$becoming a core benchmark. As of the close on February 13, 2026, $KNOWLEDGE ATLAS (02513.HK)$Reported at HK$485 per share, with a monthly increase of 114.41%, market capitalization has risen to HK$216.2 billion, continuing to break the historical high since its listing.

$KNOWLEDGE ATLAS (02513.HK)$The direct driving factor behind the surge in stock price is the commercialization of its core products: the company recently increased the subscription plan price for its artificial intelligence coding service GLM Coding Plan, with an overall increase starting at 30%. Specific adjustments include eliminating the first purchase discount while retaining quarterly and yearly subscription discounts; prices for existing subscribers remain unchanged. The new pricing will take effect on February 12, 2026. This price adjustment is not a blind increase but is based on the continued strong growth in product market demand—recently, the user base and usage of GLM Coding Plan have rapidly increased. To ensure service stability and quality under high load, the company has increased investment in computing power and model optimization, continuously upgrading product capabilities. The essence of this price adjustment is the reasonable realization of the product’s commercial value.

On a deeper level, $KNOWLEDGE ATLAS (02513.HK)$the company’s technical strength and commercialization capabilities provide solid support for its stock price. The company originated from the technology transfer of Tsinghua University's Knowledge Engineering Laboratory (KEG), with R&D personnel accounting for 74% of the workforce, and its core team possesses profound academic expertise. The GLM series models are rapidly iterating—on February 12, 2026, GLM-5 was launched simultaneously, achieving state-of-the-art open-source model scores in multiple mainstream benchmarks for programming abilities and agent performance. Its proxy programming test results rank among the top globally, surpassing Claude Opus 4.6, which was released by Claude in February. Additionally, it has completed deep integration with major domestic chip platforms such as Huawei Ascend and Moore Threads. Financial data shows that $KNOWLEDGE ATLAS (02513.HK)$revenue demonstrates high growth characteristics: revenues for 2022–2024 were RMB 57 million, RMB 125 million, and RMB 312 million respectively, with a compound annual growth rate exceeding 130%; revenue for the first half of 2025 was RMB 191 million, a year-over-year increase of 325%, surpassing the full-year level of 2023. The commercialization path is gradually becoming clearer.

Figure 2: $KNOWLEDGE ATLAS (02513.HK)$Revenue situation over the past three years

$MINIMAX-WP (00100.HK)$The new M2.5 model has entered internal testing in overseas Agent products, focusing on the optimization and implementation of agent capabilities. As a core direction for large models to evolve from tools to autonomous decision-making systems, agents can independently understand tasks, plan steps, and invoke tools to fulfill complex requirements. $MINIMAX-WP (00100.HK)$Integrating the new model into internal testing with overseas Agent products demonstrates proactive exploration in the commercial implementation of agent technology, aligning with the current trend in the large model industry transitioning from 'model capability' to 'application value'.

Figure 3: $MINIMAX-WP (00100.HK)$The new M2.5 model

$BABA-W (09988.HK)$Recent upgrades in the large model field have focused on multimodality and open-source ecosystems. Firstly, the release of the Qwen-Image-2.0 model marks another evolution in image production capability, further improving the clarity, accuracy, and diversity of image generation, adapting to more image creation scenarios. Secondly, the upcoming release of the new base model Qwen 3.5 is anticipated; this model has already appeared in the HuggingFace open-source community, with related code merge requests submitted to the Transformers library. It employs a new hybrid attention mechanism and may natively support visual understanding as a multimodal large model. Alibaba plans to open source a dense model with at least 2 billion parameters and a MoE sparse model with 35 billion parameters under the A3B architecture, aiming to strengthen its competitiveness in the open-source ecosystem and multimodal AI fields.

To summarize, the recent surge in Hong Kong-listed large model stocks ‘twin giants’ $KNOWLEDGE ATLAS (02513.HK)$with record-high stock prices, alongside major domestic firms upgrading their large models, reflects the concentrated manifestation of China’s large model industry undergoing 'technological breakthrough—demand release—capital recognition'. $KNOWLEDGE ATLAS (02513.HK)$The robust performance of the stock price is not only due to core product price adjustments and GLM-5 technological breakthroughs but also stems from a clear commercialization path, high-growth financial performance, and deep technical accumulation. Meanwhile, frequent iterations by several major companies reflect that China’s large model industry has entered a critical phase of technological deepening and application implementation. Multimodal integration, specialized breakthroughs, and accelerated commercialization have become the core themes of the industry. Presently, China’s large models have achieved global leadership in certain areas (e.g., programming, long-context processing), and their commercial capabilities are gradually being validated. The strong performance of Hong Kong's AI sector represents positive capital market feedback on the industry’s development prospects. Additionally, the sustained rise of the Hong Kong AI Index indicates overall improvement in the AI sector’s vitality rather than short-term speculation on individual companies, with the industry’s long-term foundation growing increasingly solid.

Looking ahead, China’s large model industry will maintain rapid growth, presenting a development landscape characterized by 'more precise technology, broader applications, and more mature commerce.' Technologically, multimodal fusion, long-context processing, and intelligent autonomous decision-making will be key breakthrough directions. China’s large models are expected to achieve global leadership in more niche fields while continuously improving model efficiency and further reducing computing power costs, promoting technological penetration among small and medium-sized enterprises. Moreover, AI operating systems may gradually emerge, integrating multi-model capabilities and supporting complex application development. From a commercial perspective, successful monetization experiments by certain enterprises will lead more large model companies to explore differentiated profit models, with MaaS platforms, scenario-specific customization, and open-source ecosystem services becoming mainstream. Overall profitability levels within the industry will gradually improve, with losses continuing to narrow. Additionally, deeper integration of large models across industries will accelerate, enriching use cases in finance, government, healthcare, manufacturing, and other sectors, becoming a core engine driving high-quality development of the digital economy. Overall, China’s large model industry has moved beyond the 'conceptual hype' stage, entering a virtuous development track of 'technology empowering value, commerce driving growth.' As technology continues to iterate and commercialization deepens, China’s large models are poised to secure a more important position in global AI competition, presenting long-term investment opportunities for the domestic large model sector. Investors should focus on core leaders in Hong Kong and A-shares: specifically targeting two scarce pure large model stocks in Hong Kong, plus $KNOWLEDGE ATLAS (02513.HK)$the commercial validation of these enterprises, more large model companies will gradually explore differentiated profit models. MaaS platforms, scenario-based customization, and open-source ecosystem services will become mainstream, with the industry’s overall profitability gradually improving and losses narrowing. At the same time, deeper integration of large models with various industries will accelerate, with increasing application scenarios in finance, government affairs, healthcare, manufacturing, and other fields, becoming a core engine driving high-quality development of the digital economy. Overall, China’s large model industry has shed the 'concept-hype' phase, entering a virtuous development trajectory where 'technology empowers value, commerce drives growth.' As technology iterates and commercialization deepens, China’s large models are poised to gain a more significant position in global AI competition, offering long-term development opportunities for domestic large model sectors. Investors should closely monitor core leaders in Hong Kong shares and A-shares: focusing on two rare pure large model stocks in Hong Kong, combined with $KNOWLEDGE ATLAS (02513.HK)$、 $MINIMAX-WP (00100.HK)$Two major pure large model scarce targets, superimposed. $TENCENT (00700.HK)$ 、 $BABA-W (09988.HK)$ Two major ecological giants; preferred A-share stocks $Iflytek Co.,ltd. (002230.SZ)$ Leading in the vertical implementation of the Spark large model, $Dawning Information Industry (603019.SH)$ As the cornerstone of domestic AI computing power, its 30,000-card ScaleX super cluster is the largest operational domestic computing power pool, supporting trillion-parameter model training and compatible with over 400 mainstream large models. Paired with the OneScience scientific large model development platform, it fully empowers domestic large model R&D and commercialization. The aforementioned companies cover foundational R&D, computing power support, and scenario implementation, making them the most valuable core targets for allocation in the domestic large model industry chain.