Elon Musk: China's large model will catch up with Anthropic Fable by Q1 2027! Zhipu Tangjie: It won't take that long.

Tesla CEO Elon Musk responded on X Thursday to the discussion about "When will Chinese large models catch up to Anthropic's Fable level" by providing a timeline: "Possibly in Q1 2027." Zhipu AI founder Tang Jie immediately retorted, "It won't take that long."
(Background: Anthropic released Claude Fable 5 and Claude Mythos 5 models)
(Additional context: Global AI large model usage hits 44.6 trillion tokens, with Chinese models ranking first for seven consecutive weeks)

Key Summary

• Musk estimated on X on June 18 that Chinese large models would catch up to Anthropic's Fable level around Q1 2027
• Zhipu AI founder Tang Jie replied, "It won't take that long," implying an earlier achievement
• Musk emphasized that beyond benchmarks, Anthropic's focus on "practical intelligence" will be reflected in revenue

Tesla CEO Elon Musk has also joined the conversation on X. On Thursday, when a user asked, "When will Chinese large models reach the level of Anthropic's Fable," Musk gave an estimate: "Possibly in Q1 2027."

This statement quickly prompted a response from one of the involved parties, Zhipu AI founder Tang Jie, who directly challenged Musk: "It won't take that long."

Musk replied again, elaborating further: "In benchmark tests, yes; but if measured by real-world practicality, even Q1 would be very impressive."

Unknowingly, these two AI model company founders have predicted the pace of progress for Chinese and U.S. models.

Beating benchmarks doesn't mean truly winning

Musk then added a more critical judgment. He said that Anthropic is correctly focusing on maximizing "practical intelligence," which won't be reflected in benchmark scores but will definitely show up in revenue.

Anthropic is correctly focusing on maximizing practical intelligence, which won't be reflected in benchmark scores but will definitely be reflected in revenue.

In other words, Musk is shifting the evaluation criteria from leaderboard scores to financial reports. Whether a model can top benchmarks is one thing; whether it can make people willing to pay for it is another.

Zhipu GLM-5.2 just launched

The background of this conversation is that both China and the U.S. have new developments. Anthropic's Claude Fable 5 was released publicly on June 9, representing the company's most advanced Mythos series, leading in almost all AI benchmarks, and surpassing 80% on the SWE-bench Pro, which measures real engineering capability.

On the Chinese side, Zhipu announced its flagship model GLM-5.2 on June 17, featuring 1 million tokens of lossless long context, enhanced coding ability, and support for the domestic computing platform Day0. Tang Jie's confidence in saying "it won't take that long" stems from this, and Zhipu's stock soared, giving him considerable confidence.

Frequently Asked Questions

When does Musk think Chinese large models will catch up to Anthropic's Fable?

Musk estimated on June 18 on X that it would be around Q1 2027. But he added that this refers to benchmark tests; if measured by real-world practicality, Chinese models might reach an impressive level even earlier. Zhipu founder Tang Jie responded, "It won't take that long."

What is the level of Anthropic's Claude Fable?

Fable 5 was released on June 9, representing Anthropic's most advanced Mythos series model, leading in almost all AI benchmarks. It surpasses 80% on SWE-bench Pro, which measures real engineering ability, and is regarded as one of the strongest models currently, serving as a benchmark for Chinese large models to catch up.