Video-Arena - Agentic Codex doudizhu: leaderboard
Video-Arena: 180 cases, 20 per game. Strict tuple F1 requires every non-temporal field to match and timestamps within 3 seconds; overall is the mean of nine game means. Paper v1 Appendix B.3 fixes high effort for OpenAI, Anthropic and Gemini; Kimi/MiniMax use unspecified provider defaults. Qwen thinking modes remain distinct. Full MP4, Codex harness, at most 200 turns and 120 minutes per case. Metric: Strict tuple F1 (%).
Metric: Strict tuple F1 (%). Source: videoarena-bench.github.io. Saturation forecast: Around 2030. 2 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | GPT-5.6 Sol (High) | 21.1 | #19 (GPT-5.6 Sol) |
| 2 | GPT-5.6 Luna (High) | 6.8 | #76 (GPT-5.6 Luna) |
No result here: #5 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=video-arena-agentic-codex-doudizhu · How It Works · Data refreshed daily, snapshot 2026-10-26.