Calibration
Toolathlon-Verified → GDP.pdf
GDP.pdf is estimated from Toolathlon-Verified with a offset logistic curve fitted on 7 models measured on both: y = 0.1486 + (0.3118 − 0.1486) / (1 + exp(−200.00·(x − 0.7741))), R² = 0.87, cross-validated error 3.2 pp. It is used for 5 estimates.
| Estimated model | Toolathlon-Verified | GDP.pdf | Source |
|---|---|---|---|
| Claude Opus 5 | 80.6% | 31.2% | estimated ± 3.2 pp, low confidence |
| GLM-5.3-Flash | 78.4% | 29.2% | estimated ± 3.2 pp, low confidence |
| DeepSeek V4 Pro 0813 | 74.1% | 14.9% | estimated ± 3.2 pp, medium confidence |
| Ornith-1.5-397B | 71.2% | 14.9% | estimated ± 3.2 pp, low confidence |
| DeepSeek V4 Flash 0731 | 70.3% | 14.9% | estimated ± 3.2 pp, low confidence |
Fitted on
| Model | Toolathlon-Verified | GDP.pdf |
|---|---|---|
| Claude Fable 5.1 | 77.8% | 26.2% |
| Claude Opus 5.5 | 77.8% | 26.2% |
| Claude Sonnet 5.5 | 77.8% | 25.8% |
| GLM-5.3 | 73.0% | 11.2% |
| MiMo-V2.6-Pro | 76.9% | 19.2% |
| Muse Spark Open | 72.8% | 18.6% |
| Step 5 Preview | 74.1% | 14.8% |