CadQueryEval
Code · 2026-01-10
CadQueryEval measures the percentage of 25 natural-language CAD tasks for which a model's one-shot CadQuery Python program executes, exports an STL, and passes every geometry check against a reference model: watertightness, component count, bounding-box dimensions, volume, Chamfer distance, and 95th-percentile Hausdorff distance.
Top models (higher is better)
| Model | Score |
|---|---|
| GPT-5.6 Sol Pro | 84.0 |
| Gemini 3.1 Pro Preview | 80.0 |
| GPT-5.6 Luna | 76.0 |
| GPT-5.6 Luna Pro | 76.0 |
| GPT-5.6 Sol | 76.0 |
| Kimi K3 | 76.0 |
| GPT-5.6 Terra Pro | 72.0 |
| Qwen3.7-Max | 72.0 |
| Claude Fable 5 | 64.0 |
| GPT-5.5 | 64.0 |
| Opus 4.6 | 60.0 |
| Sonnet 5 | 60.0 |
| Gemini 3.5 Flash | 60.0 |
| Grok 4.3 | 60.0 |
| Grok 4.5 | 60.0 |
| Opus 4.8 | 56.0 |
| GLM-5.1 | 56.0 |
| GPT-5.6 Terra | 56.0 |
| Kimi K2.6 | 56.0 |
| Muse Spark 1.1 | 56.0 |
| Opus 4.7 | 52.0 |
| GPT-5 Mini | 52.0 |
| Hy3 preview | 52.0 |
| Qwen3.7-Plus | 52.0 |
| Gemini 3 Pro Preview | 48.0 |
| Kimi K2.5 | 48.0 |
| Sonnet 4.5 | 44.0 |
| GLM-5.2 | 44.0 |
| MiniMax M3 | 44.0 |
| O1 | 44.0 |
| Opus 4.5 | 40.0 |
| Gemini 3 Flash Preview | 40.0 |
| DeepSeek-V4-Flash | 36.0 |
| Gemini 3.1 Flash-Lite | 36.0 |
| GPT-5 | 36.0 |
| GPT-5.1 | 36.0 |
| GPT-5.2 | 36.0 |
| GPT-5.4 | 36.0 |
| o3 | 36.0 |
| Sonnet 4.6 | 32.0 |