Musk Predicts Chinese AI Will Reach Fable Level by Q1 2027; Zhipu AI's Tang Jie Responds: 'It Won't Take That Long'

Recently, Tesla CEO Elon Musk and Zhipu AI Chief Scientist Tang Jie engaged in a public discussion on X platform regarding when Chinese AI models will reach the level of Anthropic's Fable/Mythos. Musk predicted that Chinese AI would achieve this by the first quarter of 2027 (i.e., early next year), while Tang Jie responded, "It won't take that long." This debate, set against the backdrop of the GLM-5.2 release and U.S. export controls on Anthropic's frontier models, has attracted global attention.

Event Background

• On June 9, Anthropic released Claude Fable 5 and Mythos 5, but four days later, they were taken offline due to U.S. Department of Commerce export controls, prohibiting access to Chinese users.
• On June 17, Zhipu AI released GLM-5.2 under the MIT open-source license, allowing free commercial use, supporting 1M stable context, and introducing the IndexShare mechanism to reduce computational load.
• Tech blogger @teortaxesTex analyzed that the gap between Chinese AI and U.S. frontier models is about seven months. Based on this, Musk predicted Chinese AI would reach Fable level by Q1 2027. Tang Jie promptly refuted, hinting at an earlier breakthrough.

Key Details

• GLM-5.2 achieved first place among globally available models in the Code Arena blind test and ranked among the top three globally (first among open-source models) on the Artificial Analysis comprehensive leaderboard with a score of 51.
• On benchmarks such as FrontierSWE and Terminal-Bench, the gap between GLM-5.2 and Claude Opus 4.8 narrowed to 1%–4%.
• On the BridgeBench reasoning leaderboard, GLM-5.2 defeated Fable 5 with a score of 42.8, securing the global top spot.
• Speed reaches 300 tokens/second, with cost only one-tenth that of U.S. frontier models.

Reactions

• In his response, Musk still emphasized Anthropic's advantages in "actual deployment value and revenue generation."
• Silicon Valley investment mogul Marc Andreessen reposted Tang Jie's response with the comment "Interesting," causing the event to go viral.
• Zhipu's official account responded with three "taking notes" emojis, hinting at an even earlier timeline.
• HuggingFace provided free global computing power for GLM-5.2 for six consecutive hours, the first time for a Chinese model.

Impact and Trends

• U.S. export controls have failed to halt Chinese AI progress and instead accelerated the adoption of open-source alternatives.
• The call volume of Chinese open-source models on OpenRouter has risen from 1.2% at the end of 2024 to over 50%.
• GLM-5.2 has been adapted to domestic computing platforms such as Huawei Ascend and T-Head, forming a full-stack open ecosystem.
• Analysts believe that open-source models are following a disruptive innovation path, eroding the closed-source market with cost advantages (often 90% lower) and customization.