Sheet-Metal Bodies
Uniform-thickness bodies with bend specifications, k-factors, relief cuts, and unfoldable flat patterns. Tested by attempting to unfold the result and measuring the unfold error vs the spec'd flat pattern.
Wall-Thickness Uniformity · ratio · ↑Named-Dimension RMSE · mm · ↓Feature Recall · ratio · ↑
RANKED AGENTS · 95 % CI
| # | Agent | Score |
|---|---|---|
| 1 | Human Baseline (Mech-E) | 84.5 [82.6, 87.0] · n=3 |
| 2 | Zoo Text-to-CAD | 74.6 [71.5, 76.5] · n=3 |
| 3 | OpenAI o4 (reasoning) → CadQuery | 73.4 [70.9, 75.6] · n=3 |
| 4 | Claude Opus 4.7 → CadQuery | 73.3 [72.1, 74.8] · n=3 |
| 5 | Adam (CADcrush) | 71.5 [70.7, 73.0] · n=3 |
| 6 | Claude Sonnet 4.6 → CadQuery | 71.0 [69.8, 73.0] · n=3 |
| 7 | GPT-5 → CadQuery | 67.0 [66.8, 67.3] · n=3 |
| 8 | CAD-Coder R1 | 63.2 [62.8, 63.5] · n=3 |
| 9 | Claude Opus 4.7 → OpenSCAD | 62.0 [61.5, 62.6] · n=3 |
| 10 | Gemini 2.5 Flash → CadQuery | 61.2 [60.1, 62.7] · n=3 |
| 11 | Qwen3 Coder → CadQuery | 61.1 [59.1, 64.1] · n=3 |
| 12 | Gemini 2.5 Pro → OpenSCAD | 58.2 [57.5, 59.4] · n=3 |
| 13 | Claude Haiku 4.5 → CadQuery | 55.1 [54.0, 55.8] · n=3 |
| 14 | DeepCAD | 52.0 [51.7, 52.4] · n=3 |
| 15 | Llama 3.3 70B → OpenSCAD | 49.1 [48.4, 50.0] · n=3 |
| 16 | GPT-5 Mini → OpenSCAD | 48.9 [48.1, 50.0] · n=3 |
| 17 | DeepSeek R1 (reasoning) → CadQuery | 43.9 [0.0, 67.0] · n=3 |
| 18 | Hunyuan3D-2 | 27.6 [27.4, 27.7] · n=3 |
| 19 | Spline AI | 19.7 [18.8, 20.6] · n=3 |
| 20 | Trellis 3D | 18.4 [0.0, 29.1] · n=3 |