KNOWLEDGE ATLAS (02513) soars over 40%, "Pony Alpha" ignites market expectations.

On February 9th, the enthusiasm of the capital market for domestic large models was ignited by a mysterious news, where KNOWLEDGE ATLAS (02513) surged over 40% at one point during trading, with significantly increased trading volume. This sudden market frenzy was triggered by an anonymous model codenamed "Pony Alpha" that quietly appeared on the global model service platform OpenRouter. Since its low-key launch on the OpenRouter platform on February 6th, the model quickly sparked discussions in the developer community with its outstanding programming abilities, long context understanding, and high-precision tool performance. It was widely speculated to be a preview version of the GLM-5, which KNOWLEDGE ATLAS is expected to release soon. The anonymous model sparked identity speculations. Although the provider on OpenRouter was labeled as "Stealth," and no information about its architecture, parameters, or laboratory was disclosed, this did not stop the global tech community from speculating about its identity. Within hours of the model going online, speculations about its identity were already rife. Well-known bloggers, Replit CEOs, Abacus.AICEOs, and other industry professionals joined the guessing game. Among them, the voice of "domestic large model" gradually gained the most resonance, and the viewpoint that it could be a new model from KNOWLEDGE ATLAS GLM-5 garnered the most support. These speculations were not baseless but built on a combination of multiple technological logic and industry clues. Firstly, the technical characteristics of Pony Alpha align closely with the publicly disclosed technology roadmap of KNOWLEDGE ATLAS. Pony Alpha was officially described as an "cutting-edge foundational model," emphasizing its strong performance in programming, intelligent workflow, and reasoning, with "extremely high tool call accuracy." This aligns perfectly with KNOWLEDGE ATLAS's strategic focus on code generation, agent capabilities, and advancing large models from "Chat" to "Act". Community test cases were more convincing: some developers used Pony Alpha with Claude Code to run a Minecraft project, generating 170KB of pure JavaScript code in about 2 hours, with output quality rated as "exceeding expectations". This capability to handle complex, long-term development tasks is the core competitive strength of the next generation practical large models. Secondly, the open-source clues pointed clearly towards the model's origin. Kilo Code, a partner of OpenRouter, left a crucial "metaphor" in a blog post: Pony Alpha is a "special evolution version of the most popular open-source model from a global lab." In the global mainstream open-source large model camps, Meta's Llama series and KNOWLEDGE ATLAS's GLM series are recognized as the top two. Considering the high acknowledgement of the Chinese origins of Pony Alpha and its deep optimization of intelligent workflow, the market is more inclined to believe that it is an evolution branch from GLM. In addition, KNOWLEDGE ATLAS itself has released strong signals. Not long ago, Professor Tang Jie, the chief scientist of KNOWLEDGE ATLAS, posted on social media that "GLM is coming soon," which the market interpreted as a "preview" before the release of the new model. Some developers pointed out after testing, "If it really is GLM-5, this is a clever way. Quietly collecting real feedback, while creating enough mystery and discussion hype." This strategy that combines practicality and marketing itself demonstrates a mature product mindset. KNOWLEDGE ATLAS's commercialization matrix is at a critical eruption point. If Pony Alpha is indeed a preview version of GLM-5, its significance is far beyond just technical breakthrough. This means that KNOWLEDGE ATLAS has not only overcome the engineering bottleneck of billion-level parameter models, but has also achieved systematic optimization in reasoning efficiency, context management, tool calling, and other key aspects. This is not only a technological milestone but also a solid foundation for commercialization. In its latest research report, GF SEC pointed out that KNOWLEDGE ATLAS has built a complete technical matrix with its self-developed GLM base model at its core, covering multiple modalities, intelligent agents, and code generation. It has established various revenue paths through API services, local deployment, and vertical industry solutions. It is understood that KNOWLEDGE ATLAS did not blindly follow the overseas strategy of "closed source + high price," but instead adopted an "open ecosystem + layered pricing" model. Its GLM series API has long provided free quotas, attracting a large number of small and medium-sized enterprises and independent developers. Although this strategy sacrifices some short-term income, it greatly accelerates ecosystem expansion. Data shows that as of September 2025, KNOWLEDGE ATLAS has served over 12,000 corporate clients globally, over 80 million terminal user devices, and over 45 million developers, with clients spanning high-value sectors like finance, healthcare, and manufacturing. The appearance of "Pony Alpha" in an anonymous, high-performance manner on an overseas developer platform can be seen as a further extension of its ecosystem expansion strategy. By providing tools and programming experiences that are close to or even surpassing the international leading teams, KNOWLEDGE ATLAS is expected to strengthen the technical recognition of "domestic models can also handle complex tasks" among key developer groups, paving the way for the commercial transformation of the official version of GLM-5. GF SEC expects that as the model's capabilities improve and the customer base expands, KNOWLEDGE ATLAS will enter a period of rapid revenue growth from 2025 to 2027, with scale effects gradually becoming apparent, and long-term profitability showing strong certainty. The enthusiastic response of the market to "Pony Alpha" is an early pricing of this expectation. Summary The "Pony Alpha" event marks a new stage in the competition of large models: the market is no longer blindly pursuing billion parameters but focusing on whether they can truly perform. If KNOWLEDGE ATLAS can leverage its first-mover advantage in the Agent era with GLM-5, it not only has the potential to reshape the landscape of domestic large models but also to occupy a unique ecological position in the global AI infrastructure. The sharp rise in KNOWLEDGE ATLAS's stock price today is essentially a value discovery driven by expectations of technological breakthroughs. It is not just a "bet" on an anonymous model but also a "vote" for the Chinese AI industry to seize the opportunity of the next generation paradigm shift. With the final unveiling of GLM-5 and the continued advancement of Agent applications, this storm initiated by "Pony" may just be beginning. The market will continue to pay attention to KNOWLEDGE ATLAS's next steps and the powerful momentum this technological transformation brings to the entire computing and semiconductor industry chain.