Musk Predicts Chinese LLMs Will Match Anthropic Fable by Q1 2027; Zhipu’s Tang Jie Says Sooner
English summary
Zhipu AI released and open-sourced its latest flagship model GLM-5.2, which scored 74.4 on the FrontierSWE coding benchmark, just 1 percentage point behind Anthropic’s top closed-source model Claude Opus 4.8. In an exchange on X, Elon Musk predicted that Chinese models would reach Anthropic Fable level by Q1 2027, while Zhipu co-founder Tang Jie responded that it would take less time. Musk later added that matching on benchmarks is relatively easy, but achieving comparable practical utility by Q1 2027 would still be impressive.
Chinese summary
智谱AI正式发布并开源新一代旗舰模型GLM-5.2，在FrontierSWE编程基准上得分74.4，仅落后Anthropic顶级闭源模型Claude Opus 4.8约1个百分点。马斯克在社交平台X上预测，中国大模型要到2027年第一季度才能赶上Anthropic Fable水平；智谱联合创始人唐杰回应称不需要那么久。马斯克随后补充，在跑分上追赶相对容易，但以实用性衡量，即使到2027年Q1达到也属出色表现。
Key points
Zhipu AI released open-source model GLM-5.2, scoring 74.4 on FrontierSWE benchmark, close to Anthropic Claude Opus 4.8.
智谱AI开源了GLM-5.2模型，在FrontierSWE基准测试中得分74.4，接近Anthropic Claude Opus 4.8。
Elon Musk predicted Chinese large models would reach Anthropic Fable level by Q1 2027.
马斯克预测中国大模型将在2027年第一季度达到Anthropic Fable水平。
Zhipu co-founder Tang Jie responded that it would not take that long, suggesting a shorter timeline.
智谱联合创始人唐杰回应称不需要那么久，暗示差距缩小更快。
Musk nuanced that matching benchmark scores is easier than achieving comparable practical utility.
马斯克补充说明，在基准测试上追赶容易，但达到同等实用性更难。