Calibration
GDP.pdf → Toolathlon-Verified
Toolathlon-Verified is estimated from GDP.pdf with a Michaelis–Menten curve fitted on 8 models measured on both: y = 0.8021·x / (0.00993 + x), R² = 0.32, cross-validated error 2.2 pp. It is used for 6 estimates.
| Estimated model | GDP.pdf | Toolathlon-Verified | Source |
|---|---|---|---|
| Claude Haiku 5.5 | 20.8% | 76.6% | estimated ± 2.2 pp, medium confidence |
| DeepSeek V4.1 Flash | 12.8% | 74.4% | estimated ± 2.2 pp, medium confidence |
| GPT-6 Luna | 22.8% | 76.9% | estimated ± 2.2 pp, medium confidence |
| Grok 4.7 | 20.0% | 76.4% | estimated ± 2.2 pp, medium confidence |
| Mistral Large 4 | 18.6% | 76.1% | estimated ± 2.2 pp, medium confidence |
| Muse Glimmer 30B | 10.0% | 73.0% | estimated ± 2.2 pp, low confidence |