VideoGAIA: leaderboard
Metric: Pass@1 accuracy (%; all 271 tasks; unified ReAct agent loop: 20 uniformly sampled frames plus web search, page visit and a thinking-with-videos tool that samples up to 20 more frames from a chosen segment, at most 40 steps; answers judged against the human-verified reference by GPT-5.5 (DeepSeek-V4-Pro fallback); mean of three runs). Source: arxiv.org. Saturation forecast: Around 2032. 20 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | Seed 2.0 Pro | 58.3 |
| 2 | Qwen 3.7 Plus | 56.09 |
| 3 | Kimi K3 | 54.61 |
| 4 | GPT-5.5 | 53.14 |
| 5 | Kimi K2.6 | 52.77 |
| 6 | GPT-5.4 | 52.03 |
| 7 | Qwen 3.5 Plus | 52.03 |
| 8 | Gemini 3.1 Pro (Preview) | 50.92 |
| 9 | GPT-5.2 | 50.92 |
| 10 | Qwen 3.5 397B A17B | 50.18 |
| 11 | Qwen 3.8 Max | 50.18 |
| 12 | GLM-5V Turbo | 50.18 |
| 13 | Kimi K2.5 | 49.82 |
| 14 | Qwen 3.6 Plus | 49.08 |
| 15 | MiMo-V2.5 | 48.71 |
Interactive version: theaggregate.ai/benchmark?slug=videogaia · How It Works · Data refreshed daily, snapshot 2026-09-26.