Calibration
BrowseComp → WideResearch
WideResearch is estimated from BrowseComp with a Michaelis–Menten curve fitted on 9 models measured on both: y = 1.4235·x / (0.66355 + x), R² = 0.75, cross-validated error 4.6 pp. It is used for 20 estimates.
| Estimated model | BrowseComp | WideResearch | Source |
|---|---|---|---|
| Agents-A1-4B | 66.8% | 71.4% | estimated ± 4.6 pp, high confidence |
| Claude Mythos 5 | 88.0% | 81.2% | estimated ± 4.6 pp, high confidence |
| Claude Opus 4.8 | 84.3% | 79.7% | estimated ± 4.6 pp, high confidence |
| GLM-4.7 | 52.0% | 62.5% | estimated ± 4.6 pp, medium confidence |
| GLM-5.1 | 68.0% | 72.0% | estimated ± 4.6 pp, high confidence |
| GPT-5.4 Pro | 89.3% | 81.7% | estimated ± 4.6 pp, high confidence |
| GPT-5.5 Pro | 90.1% | 82.0% | estimated ± 4.6 pp, high confidence |
| GPT-5.6 Luna | 83.3% | 79.2% | estimated ± 4.6 pp, high confidence |
| GPT-5.6 Sol | 92.2% | 82.8% | estimated ± 4.6 pp, high confidence |
| GPT-5.6 Terra | 87.5% | 81.0% | estimated ± 4.6 pp, high confidence |
| Inkling | 77.1% | 76.5% | estimated ± 4.6 pp, high confidence |
| Kimi K2.5 (Reasoning) | 60.6% | 67.9% | estimated ± 4.6 pp, high confidence |
| LongCat-Flash-Lite-Sparse | 48.6% | 60.2% | estimated ± 4.6 pp, medium confidence |
| MiniMax M3 | 83.5% | 79.3% | estimated ± 4.6 pp, high confidence |
| Nemotron 3.5 Lightning 30B A3B NVFP4 | 36.8% | 50.8% | estimated ± 4.6 pp, medium confidence |
| Qwen3.5-122B-A10B | 63.8% | 69.8% | estimated ± 4.6 pp, high confidence |
| Qwen3.5-27B | 61.0% | 68.2% | estimated ± 4.6 pp, high confidence |
| Qwen3.5-35B-A3B | 61.0% | 68.2% | estimated ± 4.6 pp, high confidence |
| Beam | 77.4% | 76.6% | estimated ± 4.6 pp, high confidence |
| Solar Pro 4 | 49.2% | 60.6% | estimated ± 4.6 pp, medium confidence |