benchgap
Calibration

WideResearch → MCP-Tasks

MCP-Tasks is estimated from WideResearch with a Hill curve fitted on 5 models measured on both: y = 0.0000 + (1.1307 − 0.0000)·x^6.00 / (0.68509^6.00 + x^6.00), R² = 0.54, cross-validated error 6.0 pp. It is used for 5 estimates.

Estimated modelWideResearchMCP-TasksSource
Atria Dawn Preview81.9%84.2%estimated ± 6.0 pp, low confidence
Qwen3.8 Max81.9%84.2%estimated ± 6.0 pp, low confidence
Ornith-1.5-397B80.8%82.4%estimated ± 6.0 pp, low confidence
dots3-note Preview78.9%79.1%estimated ± 6.0 pp, low confidence
Ornith-1.5-35B-A3B67.8%54.8%estimated ± 6.0 pp, low confidence

Fitted on

ModelWideResearchMCP-Tasks
Claude Opus 4.576.4%71.8%
GLM-569.8%60.8%
Kimi K2.572.7%59.1%
Qwen3.5 397B74.0%74.2%
Qwen3.6 Plus74.3%74.1%