Calibration
GDPval-AA → BrowseComp
BrowseComp is estimated from GDPval-AA with a Michaelis–Menten curve fitted on 28 models measured on both: y = 1.1061·x / (0.15563 + x), R² = 0.59, cross-validated error 10.1 pp. It is used for 23 estimates.
| Estimated model | GDPval-AA | BrowseComp | Source |
|---|---|---|---|
| A.X K2 | 23.2% | 66.2% | estimated ± 10.1 pp, low confidence |
| DeepSeek V3 0324 | 0.0% | 0.0% | estimated ± 10.1 pp, low confidence |
| Gemma 4 12B | 0.0% | 0.0% | estimated ± 10.1 pp, low confidence |
| Gemma 4 26B A4B | 3.4% | 19.8% | estimated ± 10.1 pp, low confidence |
| Gemma 4 E2B | 0.0% | 0.0% | estimated ± 10.1 pp, low confidence |
| Gemma 4 E4B | 0.0% | 0.0% | estimated ± 10.1 pp, low confidence |
| GPT-4.1 mini | 0.0% | 0.0% | estimated ± 10.1 pp, low confidence |
| GPT-4.1 nano | 0.0% | 0.0% | estimated ± 10.1 pp, low confidence |
| GPT-4o | 0.0% | 0.0% | estimated ± 10.1 pp, low confidence |
| GPT-4o mini | 0.0% | 0.0% | estimated ± 10.1 pp, low confidence |
| Granite 4.2 30B | 3.9% | 22.2% | estimated ± 10.1 pp, low confidence |
| Granite 4.2 3B | 0.0% | 0.0% | estimated ± 10.1 pp, low confidence |
| K-Exaone | 0.0% | 0.0% | estimated ± 10.1 pp, low confidence |
| Ling 2.6 Flash | 0.0% | 0.0% | estimated ± 10.1 pp, low confidence |
| Ling 3.0 Flash VL | 33.2% | 75.3% | estimated ± 10.1 pp, low confidence |
| Ling 3.0 Tiny | 3.2% | 18.9% | estimated ± 10.1 pp, low confidence |
| MiMo-V2-Flash | 6.6% | 32.9% | estimated ± 10.1 pp, low confidence |
| MiniCPM5-2B | 10.5% | 44.6% | estimated ± 10.1 pp, low confidence |
| Mistral Large 4 | 46.2% | 82.7% | estimated ± 10.1 pp, low confidence |
| Nemotron 3 Nano Omni 30B A3B | 0.0% | 0.0% | estimated ± 10.1 pp, low confidence |
| North Mini Code | 0.0% | 0.0% | estimated ± 10.1 pp, low confidence |
| Solar Pro 3 | 0.0% | 0.0% | estimated ± 10.1 pp, low confidence |
| Ultravox v0.6 Llama 3.3 70B | 0.0% | 0.0% | estimated ± 10.1 pp, low confidence |