Calibration
SimpleVQA → OmniDocBench 1.5
OmniDocBench 1.5 is estimated from SimpleVQA with a linear curve fitted on 8 models measured on both: y = 0.1633·x + 0.7817, R² = 0.43, cross-validated error 2.0 pp. It is used for 7 estimates.
| Estimated model | SimpleVQA | OmniDocBench 1.5 | Source |
|---|---|---|---|
| Step 3.7 Flash | 79.2% | 91.1% | estimated ± 2.0 pp, medium confidence |
| Muse Spark Open | 73.6% | 90.2% | estimated ± 2.0 pp, medium confidence |
| dots3-note Preview | 72.5% | 90.0% | estimated ± 2.0 pp, medium confidence |
| Gemini 3.1 Pro | 72.4% | 90.0% | estimated ± 2.0 pp, medium confidence |
| Muse Spark | 71.3% | 89.8% | estimated ± 2.0 pp, medium confidence |
| GPT-5.4 | 61.1% | 88.1% | estimated ± 2.0 pp, medium confidence |
| Grok 4.20 | 57.4% | 87.5% | estimated ± 2.0 pp, medium confidence |
Fitted on
| Model | SimpleVQA | OmniDocBench 1.5 |
|---|---|---|
| Claude Opus 4.5 | 65.7% | 87.7% |
| Gemini 3 Pro | 73.2% | 88.5% |
| GPT-5.2 | 55.8% | 85.7% |
| Kimi K2.5 | 71.2% | 88.8% |
| Qwen3.5 397B | 67.1% | 90.8% |
| Qwen3.6-35B-A3B | 58.9% | 89.9% |
| Qwen3.7 Plus | 81.7% | 91.4% |
| Qwen3.8 Max | 75.0% | 92.1% |