Calibration
MMMU-Pro → PerceptionBench
PerceptionBench is estimated from MMMU-Pro with a linear curve fitted on 5 models measured on both: y = 2.9882·x + -1.8258, R² = 0.85, cross-validated error 2.6 pp. It is used for 34 estimates.
| Estimated model | MMMU-Pro | PerceptionBench | Source |
|---|---|---|---|
| GPT-6.1 Terra | 85.5% | 72.9% | estimated ± 2.6 pp, low confidence |
| Gemini 3.1 Pro | 83.9% | 68.1% | estimated ± 2.6 pp, medium confidence |
| Gemini 3.5 Flash | 83.6% | 67.2% | estimated ± 2.6 pp, medium confidence |
| GPT-5.6 Sol | 83.0% | 65.4% | estimated ± 2.6 pp, medium confidence |
| Seed 2.1 Pro | 81.6% | 61.3% | estimated ± 2.6 pp, medium confidence |
| GPT-5.4 | 81.2% | 60.1% | estimated ± 2.6 pp, medium confidence |
| GPT-5.5 | 81.2% | 60.1% | estimated ± 2.6 pp, medium confidence |
| Gemini 3 Pro | 81.0% | 59.5% | estimated ± 2.6 pp, medium confidence |
| GPT-5.6 Terra | 80.7% | 58.6% | estimated ± 2.6 pp, medium confidence |
| Muse Spark | 80.4% | 57.7% | estimated ± 2.6 pp, medium confidence |
| Seed 2.1 Turbo | 80.1% | 56.8% | estimated ± 2.6 pp, medium confidence |
| GPT-5.2 | 79.5% | 55.0% | estimated ± 2.6 pp, medium confidence |
| Kimi K2.6 | 79.4% | 54.7% | estimated ± 2.6 pp, medium confidence |
| Qwen3.5 397B | 79.0% | 53.5% | estimated ± 2.6 pp, low confidence |
| Qwen3.7 Plus | 79.0% | 53.5% | estimated ± 2.6 pp, low confidence |
| Qwen3.6 Plus | 78.8% | 52.9% | estimated ± 2.6 pp, low confidence |
| Hy4 | 78.6% | 52.3% | estimated ± 2.6 pp, low confidence |
| Kimi K2.5 | 78.5% | 52.0% | estimated ± 2.6 pp, low confidence |
| GPT-5.6 Luna | 78.4% | 51.7% | estimated ± 2.6 pp, low confidence |
| Grok 4.3 | 78.1% | 50.8% | estimated ± 2.6 pp, low confidence |
| MiniMax M3 | 78.1% | 50.8% | estimated ± 2.6 pp, low confidence |
| Pareto 26.9 | 78.0% | 50.5% | estimated ± 2.6 pp, low confidence |
| MiMo-V2.5 | 77.9% | 50.2% | estimated ± 2.6 pp, low confidence |
| Claude Opus 4.6 | 77.3% | 48.4% | estimated ± 2.6 pp, low confidence |
| Gemma 4 31B | 76.9% | 47.2% | estimated ± 2.6 pp, low confidence |
| Qwen3.5-122B-A10B | 76.9% | 47.2% | estimated ± 2.6 pp, low confidence |
| GPT-5.4 mini | 76.6% | 46.3% | estimated ± 2.6 pp, low confidence |
| Step 5 Preview | 76.0% | 44.5% | estimated ± 2.6 pp, low confidence |
| Hy4 flash | 75.8% | 43.9% | estimated ± 2.6 pp, low confidence |
| Qwen3.6-27B | 75.8% | 43.9% | estimated ± 2.6 pp, low confidence |
| Qwen3.6-35B-A3B | 75.3% | 42.4% | estimated ± 2.6 pp, low confidence |
| Grok 4.20 | 75.2% | 42.1% | estimated ± 2.6 pp, low confidence |
| Qwen3.5-35B-A3B | 75.1% | 41.8% | estimated ± 2.6 pp, low confidence |
| Qwen3.5-27B | 75.0% | 41.5% | estimated ± 2.6 pp, low confidence |
Fitted on
| Model | MMMU-Pro | PerceptionBench |
|---|---|---|
| dots3-note Preview | 79.1% | 53.4% |
| Inkling 2 Small | 81.6% | 64.7% |
| Kimi K3.1 Pro | 84.1% | 68.3% |
| Kimi K3 | 81.6% | 58.5% |
| Qwen3.8 Max | 82.3% | 63.5% |