Arena Leaderboards

Explore model performance across tasks, including user preferences and agent capabilities.

Snapshot: 2026-11-02Source: Arena

Agent7 leaderboardsPreview: Confirmed Success1Claude Opus 5.5 (Max)19.8%2Claude Sonnet 5.5 (Max)19.8%3Gemini 4 Carbon (High)17.8%4GPT-6 Astra (Max)13.7%5Gemini 4 Argon (High)12.7%Snapshot 2026-11-02View 6 metrics