Calibration

WideResearch → OSWorld-Verified

Name: benchgap: LLM benchmark scores, measured and estimated
Creator: benchgap
License: https://creativecommons.org/licenses/by-nc/4.0/

OSWorld-Verified is estimated from WideResearch with a Hill curve fitted on 5 models measured on both: y = 0.0934 + (1.2000 − 0.0934)·x^6.00 / (0.74507^6.00 + x^6.00), R² = 0.80, cross-validated error 6.6 pp. It is used for 0 estimates.

Fitted on

Model	WideResearch	OSWorld-Verified
Claude Opus 4.5	76.4%	66.3%
Kimi K2.6	80.8%	73.1%
Kimi K2.5	72.7%	63.3%
Qwen3.5 397B	74.0%	62.2%
Qwen3.8 Max	81.9%	86.1%