Claude Fable 5 Banned While GLM‑5.2 Opens to All Users

I'm 小 G, and I noticed that Claude Fable 5 was recently blocked, while GLM‑5.2 was made fully available to all GLM Coding Plan users.

According to Anthropic’s official statements and X posts, the U.S. government cited national‑security and export‑control reasons and required Anthropic to suspend both Fable 5 and Mythos 5 for all foreign users. The definition of “foreign” goes beyond IP location; even users physically in the U.S. who do not meet the criteria, as well as Anthropic’s own foreign staff, are blocked.

The community reacted with frustration, noting the policy’s absurdity and sharing screenshots of the announcements.

Fable 5 is a crucial model for Anthropic. Prior releases such as Opus 4.7 and Opus 4.8 received lukewarm feedback, with many heavy users feeling the models regressed. By contrast, Fable 5 delivers noticeably stronger performance; the author describes it as restoring the “Claude returns” feeling, especially on complex tasks like code editing, where it examines context, respects interface boundaries, and makes sensible trade‑offs.

The AI field is highly competitive: Claude remains strong, OpenAI is catching up, and Google, xAI, Zhipu, DeepSeek, Kimi and others are all actively developing large‑language models, so no single vendor can dominate.

Zhipu announced that GLM‑5.2 is now open to all Lite, Pro, Max, and team editions under the GLM Coding Plan. The API will launch next week, the model will be open‑sourced under the MIT license the same week, and it now supports a genuine 1 million‑token context window, targeting long‑range tasks and coding scenarios.

The author interprets this swift rollout as a direct response to the Claude Fable 5 ban: rather than a vague teaser or limited beta, Zhipu delivers a concrete product with clear subscription access, an API timeline, and an open‑source schedule, effectively saying “you close your door, we open ours.”

Having tested GLM‑5.2 for less than an hour, the author notes that GLM‑5.1 is already very usable for coding and ranks in the first tier. The ultimate performance of GLM‑5.2 will become clear once developers run real projects on it.

On valuation, Zhipu’s market cap recently surged past HK$6,500 billion, roughly double JD.com’s HK$3,157 billion, illustrating the lofty valuations in the large‑model sector. Nevertheless, developers care more about usability, stability, and cost; they will keep models that get the job done and switch away from those that do not.

In summary, the rapid GLM‑5.2 release illustrates how model providers react to regulatory shocks, and developers will ultimately decide the winners through practical adoption.