Calibration
OSWorld 2.0 → AA Tau3 Banking
AA Tau3 Banking is estimated from OSWorld 2.0 with a Michaelis–Menten curve fitted on 5 models measured on both: y = 0.5392·x / (0.10274 + x), R² = 0.92, cross-validated error 14.6 pp. It is used for 3 estimates.
| Estimated model | OSWorld 2.0 | AA Tau3 Banking | Source |
|---|---|---|---|
| Claude Opus 4.7 | 13.9% | 31.0% | estimated ± 14.6 pp, low confidence |
| Claude Sonnet 4.6 | 8.3% | 24.1% | estimated ± 14.6 pp, low confidence |
| Qwen3.8 Max | 19.4% | 35.3% | estimated ± 14.6 pp, low confidence |