RealCADBench - Judge: leaderboard
RealCADBench judge score of the executed CAD part against the verified reference, conditional on the evaluator artifacts being available. Higher is better. Source: arxiv.org/abs/2609.03773.
Metric: Judge Score (%, conditional). Source: arxiv.org. Saturation forecast: Around February 2027. 9 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | GPT-5.5 | 73.79 | #33 |
| 2 | Gemini 3.1 Pro (Preview) | 73.33 | #64 |
| 3 | Kimi K3 | 72.23 | #35 |
| 4 | Claude Opus 4.8 | 68.31 | #39 |
| 5 | GPT-5.4 | 61.19 | #83 |
| 6 | Seed 2.0 Pro | 59.57 | #122 |
| 7 | Qwen 3.8 27B | 56.3 | #123 |
No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=realcadbench-judge · How It Works · Data refreshed daily, snapshot 2026-11-05.