MLVU-Test: leaderboard
MLVU-Test (BAAI, CVPR 2025) is the held-out test set of the long-video benchmark (videos of 3 minutes to 2 hours), with nine six-option multiple-choice tasks including sports QA and tutorial QA. This board is their average (M-AVG), accuracy in percent; higher is better.
Metric: Multiple-choice accuracy (%). Source: github.com. Saturation forecast: Estimated already saturated. 35 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | GPT-4o | 54.9 | #741 |
| 2 | Qwen-VL-Max | 34.2 | #1226 |
| 3 | Claude 3 Opus | 25 | #1459 |
No result here: #1 Claude Fable 5.2, #2 Gemini 4 Carbon, #5 Claude Opus 5.5.
Interactive version: theaggregate.ai/benchmark?slug=mlvu-test · How It Works · Data refreshed daily, snapshot 2026-11-12.