Musk Says Chinese AI Could Reach Fable-Level as Early as Q1 Next Year
Miles Bennett
Elon Musk on June 18 predicted China could reach Fable-level AI as soon as Q1 next year, as Zhipu's open-source GLM-5.2 closes in on Anthropic's top closed-source models and the U.S. government simultaneously restricts global access to Fable 5.
What exactly did Musk say?
Asked on X when China would reach Fable-level AI, Musk replied with two words: "Probably Q1."
This means → Musk believes the gap between the best U.S. and Chinese AI has narrowed to under six months — far shorter than the year-plus most observers had assumed.
His comment came at a moment when two things converged: a Chinese open-source model posted benchmark scores near the U.S. closed-source frontier, and U.S. export controls locked down Fable 5 access worldwide.
How strong is GLM-5.2, really?
Zhipu AI's GLM-5.2 has 753 billion parameters, ships under the MIT open-source license, and supports a stable 1-million-token context window.
On FrontierSWE — a benchmark for long-range coding ability — GLM-5.2 scored 74.4, trailing Anthropic's closed-source flagship Opus 4.8 at 75.1 and beating GPT-5.5's 72.6.
On PostTrainBench — testing an AI's ability to train smaller models — GLM-5.2 ranked second at 34.3, behind Opus 4.8's 37.2.
On the hardest benchmark, SWE-Marathon, GLM-5.2 scored 13.0 versus Opus 4.8's 26.0. In plain terms = it is nearly caught up on everyday tasks, but a meaningful gap remains on the most difficult work.
How does the community read this gap?
X user Teortaxes estimates GLM-5.2 trails frontier models by roughly seven months, placing it around the Opus 4.7–4.8 level.
Teortaxes also noted that Mythos — Anthropic's next-generation architecture — reached Preview status in early February 2026, with the full release, known as "Fable," expected around November–December 2026.
This means → if Fable ships on schedule by year-end and China's catch-up pace holds, Musk's "Q1" call is internally consistent on the timeline.
Why is the U.S. government tightening controls at the same time?
Bloomberg reported that U.S. Commerce Secretary Howard Lutnick wrote to Anthropic, citing export-control regulations, requiring the company to obtain a government license before giving any foreign person access to Fable 5 and Mythos 5.
Anthropic then shut down global access to both models on an emergency basis.
AI research group Proximal called GLM-5.2 "the first model to truly close the massive technology gap between Anthropic/OpenAI and other model providers."
This reflects a broader signal: U.S. concern over China's AI catch-up has spread from chip hardware to model software — not just restricting compute, but beginning to restrict the reachability of the models themselves.
Content is for reference only, not financial advice.