DFMMOLD-002 · DFM · Injection Mould · difficulty 4/5

Injection-mouldable enclosure half

sha256:1abf0e5d9c3b22d4…

§1Prompt verbatim

Lower half of a hand-held enclosure, 120 × 60 × 25 mm, parting line in the XY plane. Constraint set: ≥1° draft on every vertical wall, uniform 2.0 mm wall, no closed voids, four self-tapping bosses (Ø 4 mm OD, Ø 2 mm core, with gussets), shut-off ledge 0.5 mm overlap.

§2Ground-truth spec

shells1

watertighttrue

manifoldtrue

acceptance ε±0.1 mm

featuresdraft_1deg_min, wall_2.0mm_uniform, boss_x4, shutoff_ledge

§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✓

Human Baseline (Mech-E)

OpenAI + CadQuery 2.4

0.53312✗

Claude Opus 4.7 → CadQuery

Anthropic + CadQuery 2.4

0.42911✗

Gemini 2.5 Pro → OpenSCAD

Google + OpenSCAD 2024.06

0.3790✗

Claude Opus 4.7 → OpenSCAD

Anthropic + OpenSCAD 2024.06

0.3540✗

DeepCAD

Wu et al. 2021 (research)

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

Agent	Watert.	Manif.	Draft-Angle Compliance	Min-Wall Compliance	Wall-Thickness Uniformity	p50	latency	cost
Human Baseline (Mech-E)	✓	0.952	0.756	0.806	0.884	—	870.8s	$6.729
Zoo Text-to-CAD	✓	0.942	0.708	0.708	0.730	—	5.6s	$0.204
Adam (CADcrush)	✓	0.930	0.566	0.580	0.685	—	12.2s	$0.231
GPT-5 → CadQuery	✓	0.926	0.522	0.607	0.588	—	50.7s	$0.215
Claude Opus 4.7 → CadQuery	✓	0.919	0.594	0.708	0.628	—	32.0s	$0.370
Gemini 2.5 Pro → OpenSCAD	×	0.910	0.479	0.558	0.484	—	32.9s	$0.073
Claude Opus 4.7 → OpenSCAD	×	0.901	0.542	0.536	0.554	—	22.5s	$0.360
DeepCAD	×	0.887	0.354	0.340	0.400	—	4.9s	$0.023
Trellis 3D	×	0.858	0.088	0.228	0.121	—	11.4s	$0.058
Spline AI	×	0.850	0.050	0.093	0.063	—	7.6s	$0.042