Calibration
BrowseComp → ExploitGym
ExploitGym is estimated from BrowseComp with a linear curve fitted on 6 models measured on both: y = 3.2371·x + -2.5956, R² = 0.93, cross-validated error 6.1 pp. It is used for 34 estimates.
| Estimated model | BrowseComp | ExploitGym | Source |
|---|---|---|---|
| Agents-A1 | 75.5% | 0.0% | estimated ± 6.1 pp, low confidence |
| Agents-A1-4B | 66.8% | 0.0% | estimated ± 6.1 pp, low confidence |
| Atria Dawn Preview | 92.5% | 39.9% | estimated ± 6.1 pp, low confidence |
| Claude Mythos 5 | 88.0% | 25.3% | estimated ± 6.1 pp, low confidence |
| Claude Opus 4.6 | 83.7% | 11.4% | estimated ± 6.1 pp, low confidence |
| Claude Opus 4.7 (Adaptive) | 79.3% | 0.0% | estimated ± 6.1 pp, low confidence |
| Claude Opus 4.8 | 84.3% | 13.3% | estimated ± 6.1 pp, low confidence |
| Claude Opus 5 | 90.8% | 34.4% | estimated ± 6.1 pp, low confidence |
| Claude Sonnet 5 | 84.7% | 14.6% | estimated ± 6.1 pp, low confidence |
| DeepSeek V4 Flash 0731 | 73.2% | 0.0% | estimated ± 6.1 pp, low confidence |
| DeepSeek V4 Pro 0813 | 83.4% | 10.4% | estimated ± 6.1 pp, low confidence |
| dots3-note Preview | 83.3% | 10.1% | estimated ± 6.1 pp, low confidence |
| GLM-4.7 | 52.0% | 0.0% | estimated ± 6.1 pp, low confidence |
| GLM-5.1 | 68.0% | 0.0% | estimated ± 6.1 pp, low confidence |
| GPT-5.2 | 65.8% | 0.0% | estimated ± 6.1 pp, low confidence |
| GPT-5.4 Pro | 89.3% | 29.5% | estimated ± 6.1 pp, low confidence |
| GPT-5.5 Pro | 90.1% | 32.1% | estimated ± 6.1 pp, low confidence |
| Inkling-Small | 77.4% | 0.0% | estimated ± 6.1 pp, low confidence |
| Kimi K2.6 | 83.2% | 9.8% | estimated ± 6.1 pp, low confidence |
| Kimi K2.5 | 60.6% | 0.0% | estimated ± 6.1 pp, low confidence |
| Kimi K2.5 (Reasoning) | 60.6% | 0.0% | estimated ± 6.1 pp, low confidence |
| Ling 3.0 Flash | 72.2% | 0.0% | estimated ± 6.1 pp, low confidence |
| LongCat-Flash-Lite-Sparse | 48.6% | 0.0% | estimated ± 6.1 pp, low confidence |
| Nemotron 3.5 Lightning 30B A3B NVFP4 | 36.8% | 0.0% | estimated ± 6.1 pp, low confidence |
| Ornith-1.5-35B-A3B | 67.6% | 0.0% | estimated ± 6.1 pp, low confidence |
| Ornith-1.5-397B | 86.6% | 20.8% | estimated ± 6.1 pp, low confidence |
| Ornith-1.5-9B | 56.4% | 0.0% | estimated ± 6.1 pp, low confidence |
| Qwen3.5-122B-A10B | 63.8% | 0.0% | estimated ± 6.1 pp, low confidence |
| Qwen3.5-27B | 61.0% | 0.0% | estimated ± 6.1 pp, low confidence |
| Qwen3.5-35B-A3B | 61.0% | 0.0% | estimated ± 6.1 pp, low confidence |
| Qwen3.5 397B | 62.0% | 0.0% | estimated ± 6.1 pp, low confidence |
| Beam | 77.4% | 0.0% | estimated ± 6.1 pp, low confidence |
| Solar Pro 4 | 49.2% | 0.0% | estimated ± 6.1 pp, low confidence |
| Step 3.7 Flash | 75.8% | 0.0% | estimated ± 6.1 pp, low confidence |