On June 15, at the opening of the Hong Kong stock market, Zhipu's stock price rose steadily, reaching a intraday peak gain of 47.6%, setting a new daily trading record since its listing. By closing, the gain narrowed to 32.82%, with the total market capitalization exceeding HK$649.6 billion.
The immediate trigger came from two industry news items two days ago.
On June 12, Anthropic suspended access to its latest flagship models, Claude Fable 5 and Claude Mythos 5, in compliance with U.S. government export control requirements; one day later, Zhipu announced that its latest open-source model, GLM-5.2, is now available to all Coding Plan users, with API access and open-source model weights set to launch next week under the MIT license.
When the most advanced models begin to become "not necessarily available"
On June 12, Anthropic issued an official announcement stating that the U.S. government, under authority related to national security, issued an export control directive requiring the suspension of access to Claude Fable 5 and Claude Mythos 5 for all foreign nationals, affecting non-U.S. users both inside and outside the United States, including foreign employees within Anthropic.
Due to technical limitations in accurately and real-time identifying users' nationalities, Anthropic ultimately decided to temporarily disable both models for all global customers to ensure compliance. This decision came just three days after the official launch of the two models. As of now, both models remain unavailable on Anthropic’s official website, with no clear timeline for restoration.
As a proprietary model among the top performers in current capabilities, the Claude series has been deeply integrated by numerous developers and enterprises for tasks such as long-context processing, code development, and complex document handling. Its sudden service suspension has directly disrupted the workflows of many teams, causing a rapid surge in community discussions about alternative solutions.
On the day after the service suspension news spread, June 13, Zhipu announced that its open-source flagship model, GLM-5.2, is now available to all users of the Coding Plan, including Lite, Pro, Max, and Team editions, with API and model weights scheduled to launch next week under the MIT license.
Over the past few years, competition in the large model industry has primarily centered around capabilities.
Whose reasoning is stronger, whose coding skills are better, and who can break new capability boundaries first almost determines the choices of developers and enterprise customers.
However, the Anthropic incident revealed another previously overlooked issue: beyond capabilities, whether models can consistently and reliably gain access.
In this incident by Anthropic, developers and companies that heavily rely on overseas models for research and development may face the risk of sudden model unavailability, even if they have accounts and paid access.
This is why this incident sparked discussions among developers that far exceeded those typically seen with ordinary product updates.
As AI gradually evolves from a chat tool into infrastructure for software development, business operations, and even production processes, model stability, sustainability, and controllability are becoming as important as model capability.
Zhipu stated in its announcement, "Cutting-edge intelligence should not belong only to a few, nor should it be arbitrarily revoked by a handful of rules." This statement reflects the new reality currently facing the global AI industry.
From "Who Is Stronger" to "Who Is More Accessible"
The capital market's rapid response is essentially an early pricing of shifts in industry logic. Compared to stock price performance, the market is more focused on the signals emitted by GLM-5.2 itself.
According to information disclosed by Zhipu, GLM-5.2 is its most powerful open-source model to date, supporting a 1M context window and significantly enhanced capabilities for long-range coding tasks. Zhipu positions it as a "truly usable 1M context" model, aiming to address the issue of models losing context during lengthy, multi-step engineering tasks.
Among these, a core keyword is "Long Horizon Task." As AI agents evolve from conversational tools into execution tools, models must continuously process thousands of tool calls, tens of thousands of lines of code, and vast amounts of intermediate state information. The longer the context window, the stronger the model’s ability to maintain project state and task continuity.
The current industry competition has shifted from "answering questions" to "sustained work." For developers, what truly matters is not the model's parameter size, but whether it can maintain consistency and reliability over complex tasks lasting hours or even days.
Market reactions indicate that investors have clearly recognized this change.
Eastern Securities noted in its research report that the Anthropic model incident has highlighted the risk of closed-source model access being constrained by a single jurisdiction, potentially driving more enterprises to shift their core AI capabilities toward domestic foundation models and localized deployment. Meanwhile, GLM-5.2’s open-sourcing under the MIT license further lowers the barrier for enterprises to trial and integrate the model.
Over the past year, capital markets have valued large model companies primarily based on their model capabilities and market share. Today, as the global regulatory environment becomes increasingly volatile, a new valuation dimension is emerging— who can provide developers and enterprises with long-term, stable, and sustainably accessible AI capabilities.
As access to the world’s most advanced models begins to be influenced by external factors, openness, availability, and autonomous control are emerging as new leverage points in the AI competition.
This article is from the WeChat public account "GeekPark" (ID: geekpark), author: Lian Ran, editor: Zheng Xuan.