Calibration

We-Math → RealWorldQA

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

RealWorldQA is estimated from We-Math with a linear curve fitted on 6 models measured on both: y = 0.2675·x + 0.6003, R² = 0.69, cross-validated error 2.0 pp. It is used for 0 estimates.

Fitted on

Model	We-Math	RealWorldQA
Claude Opus 4.5	70.0%	77.0%
Gemini 3 Pro	86.9%	83.3%
GPT-5.2	79.0%	83.3%
Kimi K2.5	84.7%	81.0%
LFM3-VL-2B	68.6%	79.3%
Qwen3.5 397B	87.9%	83.9%