Calibration

We-Math → CharXiv

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

CharXiv is estimated from We-Math with a Michaelis–Menten curve fitted on 5 models measured on both: y = 2.0000·x / (1.27277 + x), R² = 0.66, cross-validated error 6.3 pp. It is used for 0 estimates.

Fitted on

Model	We-Math	CharXiv
Claude Opus 4.5	70.0%	68.5%
Gemini 3 Pro	86.9%	81.4%
GPT-5.2	79.0%	82.1%
Kimi K2.5	84.7%	77.5%
Qwen3.5 397B	87.9%	80.8%