RealCADBench - Execution Rate: leaderboard

Parametric CAD modelling from industrial design intents across four input regimes (text, 2D drawing, image, mesh): the share of generated CAD programs that execute. Fraction 0-1, higher is better. Source: arxiv.org/abs/2609.03773.

Metric: Execution Rate (0-1). Source: arxiv.org. Saturation forecast: Around February 2027. 9 models tracked.

Top models

#ModelScoreOverall rank
1GPT-5.50.93#33
2Claude Opus 4.80.89#39
3Gemini 3.1 Pro (Preview)0.88#64
4GPT-5.40.8#83
5Qwen 3.8 27B0.8#123
6Seed 2.0 Pro0.43#122
7Kimi K30.34#35

No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.

Interactive version: theaggregate.ai/benchmark?slug=realcadbench-execution-rate · How It Works · Data refreshed daily, snapshot 2026-11-05.