Calibration
SimpleVQA → MMMU-Pro
MMMU-Pro is estimated from SimpleVQA with a Hill curve fitted on 16 models measured on both: y = 0.0000 + (0.8130 − 0.0000)·x^6.00 / (0.36526^6.00 + x^6.00), R² = 0.94, cross-validated error 5.1 pp. It is used for 0 estimates.
Fitted on
| Model | SimpleVQA | MMMU-Pro |
|---|---|---|
| Claude Opus 4.5 | 65.7% | 70.6% |
| dots3-note Preview | 72.5% | 79.1% |
| Gemini 3.1 Pro | 72.4% | 83.9% |
| Gemini 3 Pro | 73.2% | 81.0% |
| GPT-5.2 | 55.8% | 79.5% |
| GPT-5.4 | 61.1% | 81.2% |
| Grok 4.20 | 57.4% | 75.2% |
| Kimi K2.5 | 71.2% | 78.5% |
| LFM2.5-VL-3B | 35.4% | 30.5% |
| LFM3-VL-2B | 36.3% | 45.7% |
| Muse Spark | 71.3% | 80.4% |
| Qwen3.5 397B | 67.1% | 79.0% |
| Qwen3.6-27B | 56.1% | 75.8% |
| Qwen3.6-35B-A3B | 58.9% | 75.3% |
| Qwen3.7 Plus | 81.7% | 79.0% |
| Qwen3.8 Max | 75.0% | 82.3% |