Video-MME-v2: leaderboard
Second-generation video understanding benchmark with 800 videos and non-linear consistency scoring across 3 difficulty levels. 8-option MCQ with grouped evaluation penalizing inconsistency.
Metric: Avg Accuracy w/o sub (%). Source: video-mme-v2.netlify.app. Saturation forecast: Around April 2027. 49 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Gemini 3 Pro | 56.8 | #80 |
| 2 | Doubao-Seed-2.0-Pro-260215 | 53.1 | #77 |
| 3 | Gemini 3 Flash | 52.4 | #99 |
| 4 | Kimi K2.5 | 51.2 | #144 |
| 5 | MiMo-V2-Omni | 47.1 | #253 |
| 6 | GPT-5 | 44.7 | #98 |
| 7 | Qwen 3.5 Omni Plus | 38.2 | #84 |
| 8 | Qwen 3 VL 235B A22B (Thinking) | 36.8 | #238 (Qwen 3 VL 235B A22B) |
| 9 | Gemma 4 31B | 36.7 | #200 |
| 10 | Qwen 3 VL 235B A22B Instruct | 33.8 | #273 |
| 11 | Qwen 2.5 VL 72B Instruct | 30.3 | #377 |
| 12 | Qwen 3 VL 8B (Thinking) | 29.7 | |
| 13 | Qwen 3 VL 30B A3B (Thinking) | 29.5 | #351 (Qwen 3 VL 30B A3B) |
| 14 | Qwen 3 VL 8B Instruct | 27.9 | #420 |
| 15 | Qwen3 Omni 30B A3B Instruct | 27 | #383 |
No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=video-mme-v2 · How It Works · Data refreshed daily, snapshot 2026-10-20.