Calibration
BrowseComp → AA Tau3 Banking
AA Tau3 Banking is estimated from BrowseComp with a Hill curve fitted on 6 models measured on both: y = 0.1474 + (1.2000 − 0.1474)·x^6.00 / (1.10756^6.00 + x^6.00), R² = 0.60, cross-validated error 10.2 pp. It is used for 17 estimates.
| Estimated model | BrowseComp | AA Tau3 Banking | Source |
|---|---|---|---|
| Agents-A1 | 75.5% | 24.3% | estimated ± 10.2 pp, low confidence |
| Agents-A1-4B | 66.8% | 19.6% | estimated ± 10.2 pp, low confidence |
| Atria Dawn Preview | 92.5% | 41.4% | estimated ± 10.2 pp, low confidence |
| Claude Mythos 5 | 88.0% | 35.9% | estimated ± 10.2 pp, low confidence |
| Claude Opus 4.6 | 83.7% | 31.3% | estimated ± 10.2 pp, low confidence |
| dots3-note Preview | 83.3% | 30.9% | estimated ± 10.2 pp, low confidence |
| GPT-5.2 | 65.8% | 19.2% | estimated ± 10.2 pp, low confidence |
| GPT-5.4 Pro | 89.3% | 37.4% | estimated ± 10.2 pp, low confidence |
| GPT-5.5 Pro | 90.1% | 38.4% | estimated ± 10.2 pp, low confidence |
| LongCat-Flash-Lite-Sparse | 48.6% | 15.5% | estimated ± 10.2 pp, low confidence |
| Ornith-1.5-35B-A3B | 67.6% | 19.9% | estimated ± 10.2 pp, low confidence |
| Ornith-1.5-397B | 86.6% | 34.3% | estimated ± 10.2 pp, low confidence |
| Ornith-1.5-9B | 56.4% | 16.5% | estimated ± 10.2 pp, low confidence |
| Qwen3.5-27B | 61.0% | 17.6% | estimated ± 10.2 pp, low confidence |
| Qwen3.5-35B-A3B | 61.0% | 17.6% | estimated ± 10.2 pp, low confidence |
| Qwen3.5 397B | 62.0% | 17.9% | estimated ± 10.2 pp, low confidence |
| Beam | 77.4% | 25.7% | estimated ± 10.2 pp, low confidence |