Arena.ai outlines Agent Arena category rankings across Code, Work, and Chat

Arena.ai published domain rankings for Agent Arena as of October 5, 2026, evaluating agentic problem-solving performance across Code, Work, and Chat. The rankings measure net improvements against the current frontier using live execution traces collected from real-world tasks.

Anthropic models hold the top spot in every category. Claude Fable 5.1 (Max) ranks first in Code and Work and second in Chat. Claude Sonnet 5.5 (Max) leads the Chat category, ranking fourth in Code and fifth in Work.

OpenAI's GPT 6 Astra (Max) placed in the top four across all three categories, taking second in Code and fourth in both Work and Chat. Arena.ai noted that Code and Work rankings align more closely than Chat; for instance, Gemini 4 Argon (High) ranked 14th in Code and 12th in Work, but placed fifth in Chat.