CAD-Bench
← back
PRIM-004 · Geometric Primitives · difficulty 1/5

Square-base pyramid frustum

sha256:8b701e2ddc40cf12…

§1Prompt verbatim

Solid frustum with 50 × 50 mm square base and 20 × 20 mm square top, height 30 mm, all four side faces plane. Origin at base centre. Output watertight solid.

§2Ground-truth spec

bbox50.0 × 50.0 × 30.0 mm
shells1
V−E+F2
genus0
watertighttrue
manifoldtrue
acceptance ε±0.03 mm

§3Reference render

canonical reference · drag to orbit, scroll to zoom

Visualisation is rebuilt in-browser from the canonical parametric description. Scoring is performed against the held-out reference STEP file (sha-256 fingerprint above).

§4Per-agent renders

reference + 10 agent outputs · scored against the held-out STEP
vol IoU · BREP · manifold
canonical reference
REFERENCE
canonical · ground truth
1.000100✓
Zoo Text-to-CAD
Zoo Text-to-CAD
Zoo (KittyCAD)
0.9416✓
DeepCAD
DeepCAD
Wu et al. 2021 (research)
0.8586✓
Human Baseline (Mech-E)
Human Baseline (Mech-E)
n=4 senior engineers
0.8136✓
Adam (CADcrush)
Adam (CADcrush)
CADcrush
0.80510✓
Qwen3 Coder → CadQuery
Qwen3 Coder → CadQuery
Alibaba + CadQuery 2.4
0.6778✓
CAD-Coder R1
CAD-Coder R1
CAD-Coder Labs (research)
0.67110✓
Claude Opus 4.7 → CadQuery
Claude Opus 4.7 → CadQuery
Anthropic + CadQuery 2.4
0.66511✗
Gemini 2.5 Flash → CadQuery
Gemini 2.5 Flash → CadQuery
Google + CadQuery 2.4
0.6269✓
Gemini 2.5 Pro → OpenSCAD
Gemini 2.5 Pro → OpenSCAD
Google + OpenSCAD 2024.06
0.6230✗
OpenAI o4 (reasoning) → CadQuery
OpenAI o4 (reasoning) → CadQuery
OpenAI + CadQuery 2.4
0.61411✓
Claude Opus 4.7 → OpenSCAD
Claude Opus 4.7 → OpenSCAD
Anthropic + OpenSCAD 2024.06
0.5900✗
Claude Sonnet 4.6 → CadQuery
Claude Sonnet 4.6 → CadQuery
Anthropic + CadQuery 2.4
0.58511✗
GPT-5 → CadQuery
GPT-5 → CadQuery
OpenAI + CadQuery 2.4
0.57411✗
DeepSeek R1 (reasoning) → CadQuery
DeepSeek R1 (reasoning) → CadQuery
DeepSeek + CadQuery 2.4
0.56611✗
GPT-5 Mini → OpenSCAD
GPT-5 Mini → OpenSCAD
OpenAI + OpenSCAD 2024.06
0.5319✗
Trellis 3D
Trellis 3D
Microsoft Research
0.5280✗
Claude Haiku 4.5 → CadQuery
Claude Haiku 4.5 → CadQuery
Anthropic + CadQuery 2.4
0.47814✗
Hunyuan3D-2
Hunyuan3D-2
Tencent
0.41717✗
Llama 3.3 70B → OpenSCAD
Llama 3.3 70B → OpenSCAD
Meta + OpenSCAD 2024.06
0.39914✗
no manifold solid produced
Spline AI
Spline AI
Spline.design
—2✗

Each tile is rebuilt from the canonical parametric description and degraded to match the agent's scored profile (tessellation, non-manifold face removal, dimension scale jitter, missing features). Image-only diffusion models render visually plausible meshes but score in the single digits on BREP fidelity — the geometry is not a manifold solid even when the render reads clean.

§5Per-agent metrics

ranked by Vol IoU · same data as the leaderboard, restricted to this task
AgentVol IoUBidirectional ChamferWatert.Manif.P@1p50latencycost
Zoo Text-to-CAD0.9410.081✓0.9901.000—7.7s$0.201
DeepCAD0.8580.095✓0.9730.000—5.5s$0.023
Human Baseline (Mech-E)0.8130.131✓0.9750.000—661.3s$6.926
Adam (CADcrush)0.8050.136✓0.9690.000—9.6s$0.285
Qwen3 Coder → CadQuery0.6770.141✓0.9550.000—12.8s$0.026
CAD-Coder R10.6710.108✓0.9580.000—6.6s$0.004
Claude Opus 4.7 → CadQuery0.6650.136✓0.9480.000—42.9s$0.392
Gemini 2.5 Flash → CadQuery0.6260.112✓0.9530.000—12.7s$0.018
Gemini 2.5 Pro → OpenSCAD0.6230.169✓0.9390.000—29.5s$0.091
OpenAI o4 (reasoning) → CadQuery0.6140.137✓0.9500.000—108.7s$0.958
Claude Opus 4.7 → OpenSCAD0.5900.142✓0.9430.000—25.4s$0.305
Claude Sonnet 4.6 → CadQuery0.5850.159✓0.9390.000—14.5s$0.062
GPT-5 → CadQuery0.5740.175✓0.9380.000—34.1s$0.172
DeepSeek R1 (reasoning) → CadQuery0.5660.153✓0.9300.000—120.5s$0.036
GPT-5 Mini → OpenSCAD0.5310.163✓0.9340.000—15.6s$0.009
Trellis 3D0.5280.188✓0.9270.000—14.0s$0.041
Claude Haiku 4.5 → CadQuery0.4780.164✓0.9220.000—8.5s$0.017
Hunyuan3D-20.4170.214×0.9100.000—30.6s$0.080
Llama 3.3 70B → OpenSCAD0.3990.210✓0.9130.000—18.4s$0.021
Spline AI
kernel error: BRepCheck_NotClosed
0.000—×0.0000.000—5.8s$0.038