Arena AI Model Weekly Rankings: Anthropic Dominates, Chinese Models Hold Steady

Arena released its latest weekly AI model rankings, covering June 29 to July 5, 2026. Anthropic's Claude series dominates across general, coding, and math categories. The Claude-fable-5 model tops the general chart with 1509 points. All top 10 spots are held by US companies, while Chinese models remain stable in the top 30.

In the general ranking, Alibaba's Qwen3.7-max-preview ranks 15th with 1475 points, the highest among Chinese models. Baidu's Ernie-5.1 is 27th, and Xiaomi's Mimo-v2.5-pro is 29th. For coding, Anthropic takes the top nine spots. Qwen3.7-max-preview ranks 10th, and Mimo-v2.5-pro ranks 17th. Math and long-text categories are also led by Anthropic.

Chinese models show strong performance across fields. Alibaba's Qwen3.7-max-preview is close to the global top tier. Moonshot AI's Kimi-k2.6 moved up to 13th in math. Next week, attention will be on whether new models from Google and OpenAI can shake up the rankings.