benchgap
Calibration

BrowseComp → MCP Atlas

MCP Atlas is estimated from BrowseComp with a Michaelis–Menten curve fitted on 23 models measured on both: y = 2.0000·x / (1.40267 + x), R² = 0.54, cross-validated error 10.0 pp. It is used for 7 estimates.

Estimated modelBrowseCompMCP AtlasSource
Agents-A1-4B66.8%64.5%estimated ± 10.0 pp, low confidence
Claude Mythos 588.0%77.1%estimated ± 10.0 pp, low confidence
GLM-4.752.0%54.1%estimated ± 10.0 pp, low confidence
GPT-5.4 Pro89.3%77.8%estimated ± 10.0 pp, low confidence
GPT-5.5 Pro90.1%78.2%estimated ± 10.0 pp, low confidence
Qwen3.5-27B61.0%60.6%estimated ± 10.0 pp, low confidence
Qwen3.5-35B-A3B61.0%60.6%estimated ± 10.0 pp, low confidence