Z.ai (Zhipu AI) co-founder Tang Jie declared on X on June 18 that China "won't take that long" to reach Fable-class AI capability, responding to an X user who asked his timeline after Z.ai's GLM-5.2 release. Elon Musk had estimated "Probably Q1" (of 2027) in the same thread. India Today reports an X user in the same exchange predicted "full PRC Mythos ('Fable') by Nov-Dec '26." The exchange follows a week in which the U.S. Commerce Department ordered Anthropic to suspend all foreign access to Fable 5 and Mythos 5 on June 13, citing national security concerns about a potential jailbreak technique. Anthropic called the directive "a misunderstanding," stressing the same technique works on GPT-5.5. GLM-5.2, the 744B-parameter MoE model Z.ai released the same day as the ban, ranked as the highest-scoring open-weights model on Design Arena at launch.
The X Exchange
On June 18, 2026, an X user asked Z.ai co-founder Tang Jie for his timeline for China to match "Fable class" capability, noting that GLM-5.2 "certainly shorten[ed] the gap." Elon Musk replied to the thread: "Probably Q1" -- implying Q1 2027. Tang Jie responded: "won't take that long," signalling Z.ai expects to reach Fable-class capability before early 2027, plausibly within 2026. India Today reports another user in the thread predicted "full PRC Mythos ('Fable') by Nov-Dec '26." Marc Andreessen also engaged with the post, per Dealroom.
The U.S. Export Ban
The exchange came less than a week after the U.S. Commerce Department used national security export controls to order Anthropic to suspend all foreign national access to Fable 5 and Mythos 5 on June 13, 2026 (Fortune, Al Jazeera/Reuters). The order covered users outside the U.S. and foreign national employees at Anthropic itself. The government cited concerns that a jailbreak technique -- described by Fortune as involving the phrase "Fix this code" -- could bypass Fable 5's safeguards and unlock Mythos's advanced cybersecurity capabilities.
Anthropic said it received the directive at 5:21 PM Eastern and called it "a misunderstanding." The company argued the technique was narrow, affected only one specific scenario, and produced similar outputs from other publicly available models including GPT-5.5. Anthropic said it had "no choice" but to disable both models for all users while disputing the order. Access to other Claude models, including Claude Opus 4.8, was not affected.
Fable 5 vs. Mythos 5
Al Jazeera and Fortune reporting clarifies the architecture: Fable 5 is the consumer-facing product built on Mythos technology, with its advanced cybersecurity and biotechnology capabilities blocked for end users. Mythos 5 is the more powerful underlying model, used by U.S. government agencies and selected corporate partners. Anthropic had previously described Mythos as too dangerous for broad release -- framing that Fortune notes gave the government a legal basis to invoke export controls.
GLM-5.2 and Z.ai
Z.ai released GLM-5.2 the same day as the export ban, with company messaging framing it as a response to frontier Western model restrictions. GLM-5.2 is a 744B-parameter Mixture-of-Experts system with 40B active parameters and a 1 million-token context window; Zhipu plans to release MIT-licensed open weights. The model ranked as the highest-scoring open-weights model on Design Arena at launch (Cryptobriefing). Z.ai went public in 2026 at a $6.6B valuation; Tang Jie is a Tsinghua University professor who co-founded Zhipu AI as a university spin-off in 2019.
What to Watch
Observable signals include GLM-5.2's open-weights release and independent benchmark disclosures, any Anthropic update on restoring Fable 5 and Mythos 5 access, and legislative or regulatory follow-on to the export control order. Public X exchanges signal intent but are not substitutes for verified model releases or technical publications.
Scoring Rationale #
The story ties three concurrent events -- China's GLM-5.2 launch, the unprecedented U.S. export ban on Anthropic's Fable 5 and Mythos 5, and Tang Jie's public timeline claim -- into a single geopolitical AI-race narrative that is notable for practitioners tracking frontier model availability and policy. The piece is a secondary synthesis published six days after the primary events and relies primarily on a public X exchange rather than new technical disclosures, placing it at the high end of Solid rather than clearly Notable.
Practice interview problems based on real data
1,500+ SQL & Python problems across 15 industry datasets — the exact type of data you work with.