Updated daily at 05:00 UTC
Benchmark
CadEval
Measures: Text to CAD: from a text description the model writes an OpenSCAD script that must render to a 3D object passing geometric checks against the specification.
- Maintainer
- —
- Introduced
- 2025
- Status
- Unmaintained
- Saturation
- Headroom
- Typical provenance
- Independent
- In scoring set
- No
Leaderboard
Independent and third-party results only.
| Model | Overall pass rate | Provenance | Source | Evaluated |
|---|---|---|---|---|
| o3 OpenAI · Medium effort | 74.0% | Third party | CadEval Dashboard via Epoch AI | — |
| Gemini 2.5 Pro | 64.0% | Third party | CadEval Dashboard via Epoch AI | — |
| o4-mini OpenAI · Medium effort | 62.0% | Third party | CadEval Dashboard via Epoch AI | — |
| o1 OpenAI · Medium effort | 56.0% | Third party | CadEval Dashboard via Epoch AI | — |
| Claude 3.7 Sonnet | 54.0% | Third party | CadEval Dashboard via Epoch AI | — |
| o3-mini OpenAI · Medium effort | 54.0% | Third party | CadEval Dashboard via Epoch AI | — |
| Claude 3.5 Sonnet | 48.0% | Third party | CadEval Dashboard via Epoch AI | — |
| GPT-4.1 | 42.0% | Third party | CadEval Dashboard via Epoch AI | — |
| Gemini 1.5 Pro | 34.0% | Third party | CadEval Dashboard via Epoch AI | — |
| Claude 3.5 Haiku | 32.0% | Third party | CadEval Dashboard via Epoch AI | — |
| Gemini 2.0 Flash | 30.0% | Third party | CadEval Dashboard via Epoch AI | — |
| GPT-4o | 26.0% | Third party | CadEval Dashboard via Epoch AI | — |
| GPT-4.1 mini | 16.0% | Third party | CadEval Dashboard via Epoch AI | — |
| Claude 3 Haiku | 12.0% | Third party | CadEval Dashboard via Epoch AI | — |
- Epoch AI, 'Capabilities & benchmarking', epoch.ai/benchmarks, CC BY 4.0.