cadclamp
Benchmark for AI-generated parametric CAD, scored on whether the part can actually be printed.
What it does and how it connects
CADClamp is a benchmark that scores AI-generated CAD on whether the part can actually be printed. It gives a model an engineering prompt with real dimensions and a declared process (FDM, 0.4 mm nozzle, PLA), the model returns a program in build123d, CadQuery, FreeCAD Python, OpenSCAD, Blender, Rhino 8, or Autodesk Fusion, and the engine runs it and grades the solid on wall thickness, overhangs, stability on the bed, watertightness, and the spec's dimensions.
The README's example makes the point: a model's L-bracket ran and exported a valid, watertight solid, but CADClamp scored it 0.176 because its walls were far below the minimum printable thickness and it could tip over on the bed, where a check that only runs the code would have passed it. The live leaderboard is published at the project's site, with contact sheets showing all 47 parts from one model in one language.
The code is Apache-2.0 and the benchmark data is under CDLA-Permissive-2.0. The project is at version 0.2 in development, and the README carries a caveats section.
Alternatives
Compare side by side- MARB
A benchmark that grades how correctly AI-assisted CAD places the parts of a real machine, rather than whether a single part can be printed.
- deepee
A benchmark that scores coding agents on embedded firmware, code repair, schematics, and PCB work, rather than on printable CAD.
- Text2CAD (MLYengineering)
A generator that emits STEP and FCStd artifacts from natural language, rather than a benchmark that grades generated CAD.
- constraint-kit
A generator that emits exact STEP and B-rep assemblies from plain English, rather than a benchmark that grades generated CAD.
- printwright
A chat-to-print workflow that sends the part to a Klipper printer, rather than a benchmark that grades whether generated CAD can print.
Sources
Not verified
- Pricing: none stated on the repo.
- No GitHub releases exist, so latest release fields are null.
- Real user voices: the repo has 0 open issues, and no independent user post was found.
- Related Error Index entries: no matching entry was verified.
- A named individual builder beyond the repository owner handle ChristianNHill
Last verified November 6, 2026. Listings are informational. The evidence score measures public, verifiable evidence and is not an endorsement or a compatibility claim.