On June 22, Zhipu's Hong Kong-listed stock price hit fresh highs, surging to 2,980 HKD per share intraday before closing up 15% at 2,410 HKD per share, with a market cap of 1.07 trillion HKD—breaking through the trillion-HKD milestone for the first time. The company held its annual shareholders' meeting that day to vote on submitting an A-share Science and Technology Innovation Board IPO proposal. On June 1, Zhipu announced plans to list on the A-share market and raise 15 billion RMB.

The A-share announcement initially pressured the stock price, however. Market concerns over equity dilution and the unlocking of restricted shares caused the stock to fall nearly 30% over ten trading days.

That same day, Anthropic removed its newly released Claude Fable 5 and Mythos 5 flagship models—launched less than 72 hours prior—due to U.S. government export control directives, sparking widespread discussion in AI circles. Zhipu responded swiftly, announcing via its official account on June 13 afternoon the full rollout of its latest flagship model, GLM-5.2.

The announcement opened with this statement: "At a moment when some cutting-edge models suddenly become unavailable, we choose to believe in another path: frontier intelligence should not belong to the few, nor should it be revoked at any moment by a handful of rules. It should be open, available, buildable, and serve every developer." This exchange over a trending event generated considerable discussion.

On June 15, the first trading day of that week, Zhipu released a notice confirming the launch of GLM-5.2 and projected the model would boost subsequent platform usage and API call volumes. The stock surged 32.8% that day, nearly recovering all losses from June.

Institutions including Orient Securities noted that the Anthropic incident exposed the uncontrollability of closed-source models, with "certainty premium" potentially shifting toward controllable, locally-deployed Chinese models. Zhipu's new model and open strategy came at precisely the right moment.

In AI circles, Zhipu has long been dubbed "China's version of Anthropic." From a technical standpoint, Anthropic rose to prominence through coding expertise and recently led trends through Agents—both areas Zhipu currently prioritizes. Commercially, Anthropic distinguished itself from inception by targeting B2B enterprise markets, while Zhipu became China's largest by revenue in large language models through government and enterprise clients. Both companies center their core business model on API-based "token selling" rather than traditional software sales or consumer products.

From "American mapping" to "domestic substitution," Zhipu leveraged the Anthropic incident to become a Hong Kong stock AI hotspot and even a market bellwether. Technical practitioners also noted that Zhipu's new model performance generated discussion within the community.

According to accounts, while lacking the buzz of DeepSeek, GLM-5.2's release prompted more comparative and competitive evaluations against top available flagship models—primarily OpenAI's GPT-5.5 and Anthropic's Claude Sonnet 4.6—whereas previously such benchmarking was often less organic.

Zhipu stated GLM-5.2 is its most capable open-source model to date, supporting genuinely usable 1M context windows and excelling in long-horizon tasks. The company also called it "still the strongest domestic coding model in our view."

The technical practitioners noted that overseas top models maintain absolute performance advantages, though specific task performance varies. From evaluations, GPT output proves more standardized with richer functionality, while Claude excels in detail and interaction. GLM-5.2 currently shows breakthroughs in interactive coding and complex programming, and compared with international top models, beyond lower compute consumption bringing cost-performance advantages, it demonstrates stronger understanding and reproduction of Chinese language and China-specific rules—particularly policy information—creating advantages in Chinese B2B and government local deployment.

On June 18, a user on X platform asked "When will Chinese large models reach Anthropic's Fable level?" and noted that "Zhipu's GLM-5.2 undoubtedly narrows the gap." Elon Musk, CEO of Tesla and SpaceX, replied that he estimated it would take until Q1 2027.

Zhipu founder Tang Jie promptly replied: "Won't take that long."

Musk subsequently added that from benchmark scores alone, perhaps, but if measured by practical deployed value, catching up by next Q1 would already be impressive.

This exchange sparked further tech-circle discussion covering technical routes and AI commercialization. Notably, reports indicate Anthropic's annualized revenue has surpassed OpenAI to lead the industry and is expected to achieve adjusted profitability for the first time this Q2. Zhipu posted a full-year 2025 net loss of 4.718 billion RMB, with R&D spending still 4.4 times revenue; JPMorgan previously predicted profitability in 2029. However, Chinese model development speed and differences in competitive strategy between China and the U.S. also drew attention. Some voices citing DeepSeek and Zhipu argue that Chinese model makers' long-term AGI strategy and ambition already differ from outdated stereotypes of model replication focus and short-term commercial interests.

China Science and Technology Innovation Partnership, a Zhipu shareholder, told us that when investing in Zhipu, large models weren't yet market consensus—it valued the company's solid scientific research foundation and resource endowments, and has consistently believed in its long-term development potential.

Tang Jie also stated at the AGI-Next Frontier Summit earlier this year that Zhipu would advance AGI exploration in three areas: scaling, model architecture innovation, and multimodal sensing-and-feeling fusion.

For Zhipu, the upcoming A-share IPO appears a critical node for both technical exploration and AI value deployment. According to the announcement, of the 150 billion RMB to be raised, 120 billion will fund artificial intelligence general foundation model projects, 20 billion for large model MaaS one-stop platform services, and 10 billion for working capital.

Before then, however, Zhipu faces July's restricted share unlocking and intensifying competition in China's model field. To advance beyond the trillion-HKD market cap threshold, stronger performance fundamentals remain essential.