CADEngBench - Engineering Requirements (L1): leaderboard
The engineering-requirements level of CADEngBench (MBZUAI, 2026): the percentage of 300 zero-to-CAD parts whose generated solid is valid and also meets every applicable dimension, feature-count, interface and design-for-manufacturing check measured on the re-imported B-Rep. Higher is better.
Metric: Requirement pass rate (%; share of 300 parametric part tasks (159 BenchCAD, 141 Fusion 360) whose brief, reference views, parameters and requirements the model turns into an executable CadQuery program; one response per task, no retries whose valid solid also passes every applicable dimension, feature-count, interface and design-for-manufacturing check on the re-imported B-Rep). Source: arxiv.org. Saturation forecast: Around February 2027. 8 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Gemini 3 Flash (Preview) (Minimal) | 30 | #144 (Gemini 3 Flash (Preview)) |
| 2 | Claude Sonnet 4.5 | 28.3 | #250 |
| 3 | Kimi K2.5 (Non-reasoning) | 24.3 | #256 (Kimi K2.5) |
| 4 | GPT-5.2 (2025-12-11) (Non-reasoning) | 22 | |
| 5 | GLM-4.6V (Non-reasoning) | 13.3 | #701 (GLM-4.6V) |
| 6 | Llama 4 Maverick | 12.7 | #1099 |
| 7 | Qwen 3.5 35B A3B (Non-reasoning) | 5 | #482 (Qwen 3.5 35B A3B) |
No result here: #1 Claude Fable 5.2, #2 Gemini 4 Carbon, #5 Claude Opus 5.5.
Interactive version: theaggregate.ai/benchmark?slug=cadengbench-engineering-requirements-l1 · How It Works · Data refreshed daily, snapshot 2026-11-13.