Calibration
MMMU-Pro → OCRBench V2
OCRBench V2 is estimated from MMMU-Pro with a Michaelis–Menten curve fitted on 5 models measured on both: y = 1.0251·x / (0.32293 + x), R² = 0.95, cross-validated error 4.4 pp. It is used for 42 estimates.
| Estimated model | MMMU-Pro | OCRBench V2 | Source |
|---|---|---|---|
| GPT-6.1 Terra | 85.5% | 74.4% | estimated ± 4.4 pp, low confidence |
| Kimi K3.1 Pro | 84.1% | 74.1% | estimated ± 4.4 pp, low confidence |
| Gemini 3.1 Pro | 83.9% | 74.0% | estimated ± 4.4 pp, low confidence |
| Gemini 3.5 Flash | 83.6% | 73.9% | estimated ± 4.4 pp, low confidence |
| GPT-5.6 Sol | 83.0% | 73.8% | estimated ± 4.4 pp, low confidence |
| Inkling 2 Small | 81.6% | 73.4% | estimated ± 4.4 pp, medium confidence |
| Kimi K3 | 81.6% | 73.4% | estimated ± 4.4 pp, medium confidence |
| Seed 2.1 Pro | 81.6% | 73.4% | estimated ± 4.4 pp, medium confidence |
| GPT-5.4 | 81.2% | 73.3% | estimated ± 4.4 pp, medium confidence |
| GPT-5.5 | 81.2% | 73.3% | estimated ± 4.4 pp, medium confidence |
| GPT-5.6 Terra | 80.7% | 73.2% | estimated ± 4.4 pp, medium confidence |
| Muse Spark | 80.4% | 73.1% | estimated ± 4.4 pp, medium confidence |
| Seed 2.1 Turbo | 80.1% | 73.1% | estimated ± 4.4 pp, medium confidence |
| Kimi K2.6 | 79.4% | 72.9% | estimated ± 4.4 pp, medium confidence |
| dots3-note Preview | 79.1% | 72.8% | estimated ± 4.4 pp, medium confidence |
| Qwen3.6 Plus | 78.8% | 72.7% | estimated ± 4.4 pp, medium confidence |
| Hy4 | 78.6% | 72.7% | estimated ± 4.4 pp, medium confidence |
| GPT-5.6 Luna | 78.4% | 72.6% | estimated ± 4.4 pp, medium confidence |
| Grok 4.3 | 78.1% | 72.5% | estimated ± 4.4 pp, medium confidence |
| MiniMax M3 | 78.1% | 72.5% | estimated ± 4.4 pp, medium confidence |
| Pareto 26.9 | 78.0% | 72.5% | estimated ± 4.4 pp, medium confidence |
| MiMo-V2.5 | 77.9% | 72.5% | estimated ± 4.4 pp, medium confidence |
| Claude Opus 4.6 | 77.3% | 72.3% | estimated ± 4.4 pp, medium confidence |
| Gemma 4 31B | 76.9% | 72.2% | estimated ± 4.4 pp, medium confidence |
| Qwen3.5-122B-A10B | 76.9% | 72.2% | estimated ± 4.4 pp, medium confidence |
| GPT-5.4 mini | 76.6% | 72.1% | estimated ± 4.4 pp, medium confidence |
| Step 5 Preview | 76.0% | 71.9% | estimated ± 4.4 pp, medium confidence |
| Hy4 flash | 75.8% | 71.9% | estimated ± 4.4 pp, medium confidence |
| Grok 4.20 | 75.2% | 71.7% | estimated ± 4.4 pp, medium confidence |
| Qwen3.5-35B-A3B | 75.1% | 71.7% | estimated ± 4.4 pp, medium confidence |
| Qwen3.5-27B | 75.0% | 71.7% | estimated ± 4.4 pp, medium confidence |
| Inkling-Small | 74.0% | 71.4% | estimated ± 4.4 pp, medium confidence |
| Muse Glimmer 30B | 74.0% | 71.4% | estimated ± 4.4 pp, medium confidence |
| Gemma 4 26B A4B | 73.8% | 71.3% | estimated ± 4.4 pp, medium confidence |
| Inkling | 73.5% | 71.2% | estimated ± 4.4 pp, medium confidence |
| Gemma 4 12B | 69.1% | 69.9% | estimated ± 4.4 pp, medium confidence |
| GPT-5.4 nano | 66.1% | 68.9% | estimated ± 4.4 pp, medium confidence |
| Command A+ | 63.0% | 67.8% | estimated ± 4.4 pp, medium confidence |
| Gemma 4 E4B | 52.6% | 63.5% | estimated ± 4.4 pp, medium confidence |
| Qwen2.5-VL-32B | 49.5% | 62.0% | estimated ± 4.4 pp, medium confidence |
| Mistral Small 3.1 | 49.3% | 61.9% | estimated ± 4.4 pp, medium confidence |
| Gemma 4 E2B | 44.2% | 59.2% | estimated ± 4.4 pp, medium confidence |
Fitted on
| Model | MMMU-Pro | OCRBench V2 |
|---|---|---|
| Interfaze Beta | 71.1% | 70.7% |
| LFM2.5-VL-3B | 30.5% | 47.5% |
| LFM3-VL-2B | 45.7% | 63.5% |
| Qwen3.7 Plus | 79.0% | 70.7% |
| Qwen3.8 Max | 82.3% | 74.2% |