Calibration
VideoMMMU → Video-MME (w/o subtitle)
Video-MME (w/o subtitle) is estimated from VideoMMMU with a Michaelis–Menten + offset curve fitted on 7 models measured on both: y = 0.1866 + 2.0000·x / (1.75852 + x), R² = 0.96, cross-validated error 2.3 pp. It is used for 0 estimates.
Fitted on
| Model | VideoMMMU | Video-MME (w/o subtitle) |
|---|---|---|
| Claude Opus 4.5 | 84.4% | 81.4% |
| Gemini 3 Pro | 87.6% | 87.7% |
| GPT-5.2 | 85.9% | 85.8% |
| Kimi K2.5 | 86.6% | 83.2% |
| LFM3-VL-2B | 49.6% | 62.8% |
| Qwen3.5 397B | 84.7% | 83.7% |
| Qwen3.6-35B-A3B | 83.7% | 82.5% |