benchgap
Calibration

BrowseComp → Gert Labs

Gert Labs is estimated from BrowseComp with a inverse Michaelis–Menten curve fitted on 14 models measured on both: y = 0.92586·(x − 0.0000) / (0.0000 + 2.0000 − x), R² = 0.74, cross-validated error 7.4 pp. It is used for 11 estimates.

Estimated modelBrowseCompGert LabsSource
Agents-A175.5%56.2%estimated ± 7.4 pp, medium confidence
Agents-A1-4B66.8%46.4%estimated ± 7.4 pp, medium confidence
GPT-5.4 Pro89.3%74.7%estimated ± 7.4 pp, low confidence
GPT-5.5 Pro90.1%75.9%estimated ± 7.4 pp, low confidence
Inkling77.1%58.1%estimated ± 7.4 pp, medium confidence
Inkling-Small77.4%58.5%estimated ± 7.4 pp, medium confidence
LongCat-Flash-Lite-Sparse48.6%29.7%estimated ± 7.4 pp, low confidence
Nemotron 3.5 Lightning 30B A3B NVFP436.8%20.9%estimated ± 7.4 pp, low confidence
Nemotron 3 Ultra44.4%26.4%estimated ± 7.4 pp, low confidence
Beam77.4%58.5%estimated ± 7.4 pp, medium confidence
Solar Pro 449.2%30.2%estimated ± 7.4 pp, low confidence