Calibration
BrowseComp → AA Harvey LAB
AA Harvey LAB is estimated from BrowseComp with a Michaelis–Menten + offset curve fitted on 5 models measured on both: y = 0.6842 + 2.0000·x / (6.44711 + x), R² = 0.90, cross-validated error 9.8 pp. It is used for 15 estimates.
| Estimated model | BrowseComp | AA Harvey LAB | Source |
|---|---|---|---|
| Agents-A1-4B | 66.8% | 87.2% | estimated ± 9.8 pp, low confidence |
| Atria Dawn Preview | 92.5% | 93.5% | estimated ± 9.8 pp, low confidence |
| Claude Mythos 5 | 88.0% | 92.4% | estimated ± 9.8 pp, low confidence |
| GLM-4.7 | 52.0% | 83.3% | estimated ± 9.8 pp, low confidence |
| GPT-5.2 | 65.8% | 86.9% | estimated ± 9.8 pp, low confidence |
| GPT-5.4 Pro | 89.3% | 92.8% | estimated ± 9.8 pp, low confidence |
| GPT-5.5 Pro | 90.1% | 92.9% | estimated ± 9.8 pp, low confidence |
| Kimi K2.5 | 60.6% | 85.6% | estimated ± 9.8 pp, low confidence |
| Kimi K2.5 (Reasoning) | 60.6% | 85.6% | estimated ± 9.8 pp, low confidence |
| LongCat-Flash-Lite-Sparse | 48.6% | 82.4% | estimated ± 9.8 pp, low confidence |
| Qwen3.5-27B | 61.0% | 85.7% | estimated ± 9.8 pp, low confidence |
| Qwen3.5-35B-A3B | 61.0% | 85.7% | estimated ± 9.8 pp, low confidence |
| Qwen3.5 397B | 62.0% | 86.0% | estimated ± 9.8 pp, low confidence |
| Beam | 77.4% | 89.9% | estimated ± 9.8 pp, low confidence |
| Solar Pro 4 | 49.2% | 82.6% | estimated ± 9.8 pp, low confidence |