Calibration
HLE w/ tools → WideResearch
WideResearch is estimated from HLE w/ tools with a Michaelis–Menten curve fitted on 6 models measured on both: y = 1.3403·x / (0.35505 + x), R² = 0.95, cross-validated error 3.4 pp. It is used for 16 estimates.
| Estimated model | HLE w/ tools | WideResearch | Source |
|---|---|---|---|
| Agents-A1 | 47.6% | 76.8% | estimated ± 3.4 pp, medium confidence |
| Apodex 1.1 | 56.1% | 82.1% | estimated ± 3.4 pp, medium confidence |
| Claude Haiku 5.5 | 57.4% | 82.8% | estimated ± 3.4 pp, low confidence |
| Claude Opus 5 | 64.7% | 86.5% | estimated ± 3.4 pp, low confidence |
| Claude Opus 5.5 | 67.7% | 87.9% | estimated ± 3.4 pp, low confidence |
| Claude Sonnet 5 | 57.4% | 82.8% | estimated ± 3.4 pp, low confidence |
| Claude Sonnet 5.5 | 64.5% | 86.4% | estimated ± 3.4 pp, low confidence |
| DeepSeek V4.1 Flash | 63.9% | 86.2% | estimated ± 3.4 pp, low confidence |
| DeepSeek V4 Flash 0731 | 45.1% | 75.0% | estimated ± 3.4 pp, medium confidence |
| DeepSeek V4 Pro 0813 | 60.0% | 84.2% | estimated ± 3.4 pp, low confidence |
| GLM-5.3 | 62.5% | 85.5% | estimated ± 3.4 pp, low confidence |
| GLM-5.3-Flash | 55.3% | 81.6% | estimated ± 3.4 pp, medium confidence |
| GPT-6 Astra | 57.2% | 82.7% | estimated ± 3.4 pp, low confidence |
| Nemotron 3 Ultra | 37.4% | 68.8% | estimated ± 3.4 pp, medium confidence |
| Qwen3.7 Max | 53.5% | 80.6% | estimated ± 3.4 pp, medium confidence |
| Step 3.7 Flash | 47.2% | 76.5% | estimated ± 3.4 pp, medium confidence |