Calibration
BrowseComp → AA EnterpriseOps-Gym
AA EnterpriseOps-Gym is estimated from BrowseComp with a Michaelis–Menten + offset curve fitted on 6 models measured on both: y = 0.0913 + 2.0000·x / (4.21796 + x), R² = 0.52, cross-validated error 6.9 pp. It is used for 9 estimates.
| Estimated model | BrowseComp | AA EnterpriseOps-Gym | Source |
|---|---|---|---|
| Agents-A1 | 75.5% | 39.5% | estimated ± 6.9 pp, low confidence |
| dots3-note Preview | 83.3% | 42.1% | estimated ± 6.9 pp, low confidence |
| GPT-5.4 Pro | 89.3% | 44.1% | estimated ± 6.9 pp, low confidence |
| GPT-5.5 Pro | 90.1% | 44.3% | estimated ± 6.9 pp, low confidence |
| LongCat-Flash-Lite-Sparse | 48.6% | 29.8% | estimated ± 6.9 pp, low confidence |
| Ornith-1.5-35B-A3B | 67.6% | 36.8% | estimated ± 6.9 pp, low confidence |
| Ornith-1.5-397B | 86.6% | 43.2% | estimated ± 6.9 pp, low confidence |
| Ornith-1.5-9B | 56.4% | 32.7% | estimated ± 6.9 pp, low confidence |
| Beam | 77.4% | 40.1% | estimated ± 6.9 pp, low confidence |