Loading page…
Preferred languages
More languages supported by Google Translate
Open the selector to load the current language catalog.
Translations are provided by Google Translate.
Loading page…
Preferred languages
More languages supported by Google Translate
Open the selector to load the current language catalog.
Translations are provided by Google Translate.
Multimodal & Grounded benchmark
BenchCAD Vision2Code voxel IoU with tools. Every model the catalog carries a published BenchCAD Vision2Code (tools) value for, ranked by that value.
Generates CadQuery code from multi-view renders with image inspection and code-execution tools. 6 models in the ModelScale catalog have a published value.
Each point is a model with a published BenchCAD Vision2Code (tools) value and complete token prices, on the 75% input / 25% output blend. Frontier models are spaced evenly along the cost axis.
6 models with a published BenchCAD Vision2Code (tools) value, ordered by that value, highest first. Models the source has not scored on this benchmark are not listed — they are unmeasured, not last.
| Rank | Model | Provider | Voxel IoU with tools |
|---|---|---|---|
| 1 | Anthropic | 0.963 | |
| 2 | Anthropic | 0.962 | |
| 3 | OpenAI | 0.959 |
What the benchmark measures and how its value is published, from the source’s own record.
BenchCAD Vision2Code (tools) (BenchCAD Vision2Code voxel IoU with tools) is a multimodal & grounded benchmark published by BenchCAD: A comprehensive, industry-standard benchmark for programmatic CAD. Generates CadQuery code from multi-view renders with image inspection and code-execution tools. Its evaluation set is described by the source as 1,000-file Vision2Code subset. The source rates its difficulty as programmatic CAD generation.
Each model's result is published by BenchCAD: A comprehensive, industry-standard benchmark for programmatic CAD as voxel IoU with tools. ModelScale prints that value to 3 decimal places, exactly as published, and never converts or rescales it. This page ranks the 6 models that carry a published value, highest value first; a model the source never scored is left out rather than ranked at zero. The source does not state whether a higher value is the better result, so neither does this page.
All leaderboards · Multimodal & Grounded capability leaderboard
| 4 | Moonshot AI | 0.927 |
| 5 | Anthropic | 0.843 |
| 6 | Anthropic | 0.821 |
Evidence key: Observed
Rows are ordered by the value BenchCAD: A comprehensive, industry-standard benchmark for programmatic CAD published, highest first. The source does not state whether a higher value is the better result, so this page does not either: for a benchmark that measures a rate of failure, read the table from the bottom.