multimodal benchmark
BenchCAD
BenchCAD is a benchmark for programmatic CAD reasoning built from 17,900 execution-verified CadQuery programs spanning 106 industrial part families, roughly half anchored to real ISO, DIN, EN, ASME, and IEC specification tables. It decomposes CAD capability into matched tasks; the Vision2Code task requires models to generate CadQuery code from multi-view renders, scored by voxel IoU against the reference geometry.
Updated Aug 7, 2026
Published models4
Registry coverage4
MetricScore
EvidenceB
BenchCAD leaderboard
Sorted by the source-provided rank. Higher score is better according to the registry.
| 1 | OP | 70.6% | 100.0% | 4 | C | |
| 2 | OP | 63.1% | 66.7% | 4 | C | |
| 3 | OP | 62.3% | 33.3% | 4 | C | |
| 4 | AN | 37.3% | 0.0% | 4 | C |
Score distribution
Top published rows on the benchmark's original scale.
BenchCAD highlights
The top published results on this benchmark's own scale.
What is BenchCAD?
Definition and scoring fields from the benchmark registry.
BenchCAD is a benchmark for programmatic CAD reasoning built from 17,900 execution-verified CadQuery programs spanning 106 industrial part families, roughly half anchored to real ISO, DIN, EN, ASME, and IEC specification tables. It decomposes CAD capability into matched tasks; the Vision2Code task requires models to generate CadQuery code from multi-view renders, scored by voxel IoU against the reference geometry.
Scores are shown in ratio. The current registry marks this benchmark as not independently verified with evidence level B.
- Family
- BenchCAD
- Modality
- multimodal
- Primary category
- multimodal
- Score direction
- higher
- LLMBoard eligible
- No
- Evaluation key
- benchcad|llm-stats-current
Source-native results are preserved. Eligibility for the overall LLMBoard score is a separate policy decision.
FAQ
Common questions about BenchCAD.
Which model scores highest on BenchCAD?
GPT-5.6 Sol is currently ranked first with 70.6%.
What does BenchCAD measure?
BenchCAD is a benchmark for programmatic CAD reasoning built from 17,900 execution-verified CadQuery programs spanning 106 industrial part families, roughly half anchored to real ISO, DIN, EN, ASME, and IEC specification tables. It decomposes CAD capability into matched tasks; the Vision2Code task requires models to generate CadQuery code from multi-view renders, scored by voxel IoU against the reference geometry.
Is a higher score better?
Yes. Higher values rank better for this benchmark.
How many models are compared?
4 unique published model results are currently shown.
Does this benchmark affect the overall score?
This benchmark is preserved as source-native evidence but is not eligible for the current overall score.