Video-MME — leaderboard
Comprehensive multi-modal evaluation for video understanding covering diverse video types, durations, and question formats.
Metric: Overall Accuracy (%). Source: video-mme.github.io. Status: saturation imminent. 51 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | video-SALMONN 2+THU & ByteDance | 79.7 |
| 2 | Gemini 1.5 ProGoogle | 75 |
| 3 | AdaReTaKeHIT & Huawei | 73.5 |
| 4 | Keye-VL-1.5Kwai Keye Team | 73 |
| 5 | TSPOXJTU & TeleAI & USTB | 73 |
| 6 | InternVL2.5Shanghai AI Lab | 72.1 |
| 7 | GPT-4oOpenAI | 71.9 |
| 8 | Qwen2-VLAlibaba | 71.2 |
| 9 | JT-VL-ChatChina Mobile | 70.6 |
| 10 | LLaVA-VideoBytedance & NTU S-Lab | 70.6 |
Interactive version: theaggregate.ai/benchmark?slug=video-mme · How the rankings work · Data refreshed daily, snapshot 2026-07-22.