Anthropic leads AI model rankings again, code list top changes

Arena released its latest weekly AI model rankings. Anthropic's models dominate the top spots across multiple categories including overall and coding. In the overall list, claude-fable-5 keeps first place with 1505 points, and all top five are from Anthropic, with scores within 30 points of each other.

The code list saw a change: claude-opus-4-7-thinking moved from second to first, while former leader claude-fable-5 fell to fifth. Among Chinese models, Alibaba's qwen3.7-max-preview ranks 17th overall and 12th in coding. Other Chinese models like Zhipu's glm-5.1 and Baidu's ernie-5.1 also made the list.

In the math list, claude-fable-5 rose to first with a 31-point gain. The hard instruction and multi-turn chat lists also saw minor shifts. Overall, US companies still lead, while Chinese models stay stable in the mid-to-upper range, with some climbing in coding rankings.