June 23 Hong Kong stock market data shows that large-model concept stocks overall fell: Z.ai’s intraday drop reached 14.11%, currently trading at HK$2,070; MINIMAX-W’s decline reached 11.35%, currently trading at HK$541.60. In the background, Z.ai released its latest AI model GLM-5.2, with performance continuing to outperform GPT-5.5 and Gemini 3.1 Pro.
Z.ai CEO Zhang Peng responds to Musk: China doesn’t need to wait until next year’s first quarter to benchmark Fable 5
According to a report, Elon Musk posted on X that China’s AI model targeting Anthropic’s latest Fable 5 “will probably be in the first quarter of next year.” Z.ai CEO Zhang Peng immediately replied in the comments, saying “no need for that long.”
Z.ai is one of China’s leading AI startups, and the response suggests Z.ai is developing, or is about to launch, a product that can compete with the strongest U.S. AI model. The above is Zhang Peng’s personal public statement, and no specific timeline was further detailed in the related report.
Z.ai releases GLM-5.2: benchmark performance nearly matches Anthropic Opus 4.7 and 4.8
According to a report, Z.ai released GLM-5.2 on June 16, 2026. Benchmarks show its performance is almost on par with Anthropic’s Opus 4.7 (released in April) and Opus 4.8 (released in May), while continuing to stay ahead of OpenAI’s GPT-5.5 and Google’s Gemini 3.1 Pro. The benchmarks are reported as numbers Z.ai itself submitted, with no third-party independent verification mentioned.
Background on Fable 5 export controls: made public on June 10, access banned for foreign users starting June 13
According to a report, Anthropic’s Claude Fable 5 was made public on June 10, 2026. Just three days later (June 13), the U.S. government imposed export controls on Fable 5, banning all foreigners (including Anthropic’s own employees) from accessing it. The U.S. government claimed it implemented the controls after finding that the defense mechanisms protecting Fable 5 on Amazon could be bypassed and that Anthropic refused to fix the issue; Anthropic countered that the vulnerability is minor and that it can also be reproduced on other models such as GPT-5.5.
Because it could not be ensured that the control instructions would be fully complied with, Anthropic decided to withdraw Fable 5 and Mythos 5 globally. Washington said that once the vulnerability is patched, restrictions would be lifted and foreign users would be allowed to access Fable 5 again.
FAQ
Are Z.ai and MINIMAX-W’s stock declines in Hong Kong directly related to the AI export controls?
Based on existing reports, the article does not provide specific reasons for the declines of Z.ai and MINIMAX-W in Hong Kong stocks, only stating that “Hong Kong large-model concept stocks fell.” The Fable 5 export controls and Z.ai’s GLM-5.2 release are events occurring around the same time, but the report does not explain any direct causal relationship between them and the Hong Kong stock drop.
Are the benchmark numbers for Z.ai’s GLM-5.2 independently verified?
According to the report, GLM-5.2’s benchmark performance is “almost equivalent to Anthropic Opus 4.7 and 4.8, while continuing to lead GPT-5.5 and Gemini 3.1 Pro.” This is Z.ai’s own claim at the time of release, and the report does not mention independent verification results by any third-party organization.
After the U.S.’s export controls on Fable 5, what is the current state of the AI race between China and the U.S.?
According to the report, after the U.S. government implemented export controls, Fable 5 and Mythos 5 were pulled globally by Anthropic, and the U.S. government said the restrictions would be lifted after the vulnerability is patched. On the China side, Z.ai CEO Zhang Peng said China does not need to wait until next year’s first quarter to develop a model that benchmarks Fable 5; GLM-5.2 released by Z.ai on June 16 is considered to be close to the level of Fable 5’s predecessor models Opus 4.7/4.8. At present, neither side has publicly achieved full technical parity.
