The Arena platform released its AI model rankings for the 41st week of 2026, covering the period from October 5 to October 11. Anthropic's new Claude Opus 5.5 Max debuted at the top of the agent leaderboard with a 15.53% net improvement rate, a 17.29% task confirmation success rate, and a 0.37% tool hallucination rate.
In the overall leaderboard, Gemini 4 Argon High retained the top spot with an ELO score of 1525. Anthropic models occupy four of the top positions, with Claude Opus 5.5 High rising to second place. New entries include GPT 5.5 and Gemini 3.5 Flash Medium, which debuted at 29th and 30th places respectively.
Gemini 4 Argon High continues to lead the coding category with an ELO of 1562. Claude Opus 5.5 High moved into the top five, while Zhipu GLM 5.3 Flash became the highest-ranked Chinese model in this category at 27th place. Claude Opus 5.5 Max remains the leader in the frontend development category.
Chinese models maintain a stable presence across various leaderboards. Kimi K3 Max remains in the top 20 of the overall ranking, while Xiaomi's MiMo V2.6 Flash and MiMo V2.6 Pro show progress in frontend development. In the video generation category, six Chinese models are currently ranked within the top 20.
The image generation leaderboard is dominated by OpenAI's GPT Image 2.5 Sunburst. Google's Gemini Nano Banana 2.1 entered the ranking at 6th place. ByteDance's Seedream 5.0 Pro and Alibaba's Qwen Image 3.0 Pro hold positions in the upper-middle tier of the category.
