Anthropic's Fable 5.1 Tops Benchmarks, Widening Gap with Chinese Models
nashnova research
Anthropic released Fable 5.1 on September 3, claiming the global No. 1 spot on two major AI benchmarks and leading China's top models by 6 points. This means → the performance gap between America's closed-source frontier and China's open-source camp is widening again — but on price and commercial adoption, the race is far from settled.
How strong is Fable 5.1, exactly?
Fable 5.1 scored 66 on the Artificial Analysis Intelligence Index — global No. 1, leading China's top contenders, Moonshot's Kimi K3 and Zhipu's GLM-5.3, by 6 points each.
It also topped the Vals Index — a benchmark measuring complex real-world tasks in finance, coding, and law. Anthropic's own Opus 5 and Fable 5 took second and third.
In plain terms = one company swept the top three, which means Anthropic currently has no peer on the hardest tasks.
Does the "China has caught up" narrative still hold?
Databricks engineer Yuchen Jin called Fable 5.1's capability jump "stunning."
Some observers argue the result directly breaks the narrative that Chinese open-source developers have closed the frontier gap.
This means → on the layer benchmarks actually measure — the most complex coding and knowledge work — the US-China gap is widening again, not narrowing.
But is China's real advantage somewhere else?
Chinese models take a budget-friendly, open-source route and continue gaining traction in global commercial markets.
This reflects two sharply different strategic bets: the US wagers on closed-source frontier performance; China wagers on low-cost, mass-scale adoption.
In plain terms = the US leads on "who is smartest," while China attacks on "who is most widely used" — whether frontier performance and commercial penetration can move in sync is the core question for where this competition heads.
市场有风险，内容仅供研究参考，不构成投资建议。