Calibration
Gert Labs → BrowseComp
BrowseComp is estimated from Gert Labs with a offset logistic curve fitted on 14 models measured on both: y = 0.5863 + (0.8111 − 0.5863) / (1 + exp(−53.46·(x − 0.4924))), R² = 0.84, cross-validated error 5.7 pp. It is used for 24 estimates.
| Estimated model | Gert Labs | BrowseComp | Source |
|---|---|---|---|
| Claude 4 Sonnet | 39.7% | 58.8% | estimated ± 5.7 pp, medium confidence |
| DeepSeek V3.2 | 29.6% | 58.6% | estimated ± 5.7 pp, medium confidence |
| Gemini 2.5 Pro | 42.0% | 59.1% | estimated ± 5.7 pp, medium confidence |
| Gemini 3.1 Flash-Lite | 38.5% | 58.7% | estimated ± 5.7 pp, medium confidence |
| Gemini 3 Flash | 56.6% | 80.7% | estimated ± 5.7 pp, medium confidence |
| Gemini 3 Pro | 63.2% | 81.1% | estimated ± 5.7 pp, medium confidence |
| Gemma 4 31B | 35.3% | 58.6% | estimated ± 5.7 pp, medium confidence |
| GLM-5V-Turbo | 30.8% | 58.6% | estimated ± 5.7 pp, medium confidence |
| GPT-4.1 | 25.7% | 58.6% | estimated ± 5.7 pp, low confidence |
| GPT-5.1 | 41.2% | 58.9% | estimated ± 5.7 pp, medium confidence |
| GPT-5.1-Codex | 49.7% | 71.2% | estimated ± 5.7 pp, medium confidence |
| GPT-5.2-Codex | 51.8% | 76.5% | estimated ± 5.7 pp, medium confidence |
| Grok 4 | 42.3% | 59.2% | estimated ± 5.7 pp, medium confidence |
| Grok 4.1 Fast | 47.3% | 64.6% | estimated ± 5.7 pp, medium confidence |
| Grok 4.20 | 38.4% | 58.7% | estimated ± 5.7 pp, medium confidence |
| Grok Build 0.1 | 49.2% | 69.6% | estimated ± 5.7 pp, medium confidence |
| Hy3 Preview | 36.9% | 58.7% | estimated ± 5.7 pp, medium confidence |
| MiMo-V2.5 | 46.9% | 63.6% | estimated ± 5.7 pp, medium confidence |
| MiMo-V2-Pro | 36.7% | 58.7% | estimated ± 5.7 pp, medium confidence |
| Mistral Medium 3.5 128B | 39.1% | 58.7% | estimated ± 5.7 pp, medium confidence |
| Qwen3.6-27B | 54.8% | 80.0% | estimated ± 5.7 pp, medium confidence |
| Qwen3.7 Max | 64.3% | 81.1% | estimated ± 5.7 pp, medium confidence |
| Qwen3 Max | 43.7% | 59.8% | estimated ± 5.7 pp, medium confidence |
| Trinity-Large-Thinking | 32.6% | 58.6% | estimated ± 5.7 pp, medium confidence |