partforge

Mesh rebuild freecad bracket

hard tier · create

The request

I've attached an STL of an L-bracket exported from a CAD application. Rebuild it as a clean parametric part: one sub-part named exactly `bracket`, matching the mesh's dimensions, hole positions and slots, and standing in the same place as the mesh (Z up, same origin — build it where the import sits, don't re-centre it). Give it three controls — `width`, `thickness` and `holeDiameter` — whose defaults match the mesh and whose ranges cover ±10 mm, 2–8 mm and 3–8 mm. Keep the mesh imported in the part while you work. Measure the mesh first; do not guess.
A render of the mesh the model was given
Mesh · made with FreeCAD 1.1.0 (Part workbench, MeshPart tessellation at 0.01 mm linear deflection) · CC0-1.0

What the AI judge checks

  • The overall form matches the reference: the same kind of object, the same silhouette from the matching angle
  • Proportions match: the relative sizes of the major dimensions agree with the reference within what the eye can judge
  • Every major feature visible in the reference (holes, slots, bosses, tabs, lips) is present in the result
  • Those features sit where the reference has them
  • Nothing is present that the reference does not show — no decorative or unexplained geometry
  • Edges are straight and faces are flat — a CAD rebuild, not a copy of the mesh's facets
  • The slots are elongated openings, not round holes

Measured checks

  • one sub-part named bracket (gate)
  • volume within 3 % of the mesh (gate)
  • bounding box within 0.5 mm of the mesh (gate)
  • symmetric difference under the measured floor
  • holes match the mesh
  • the imported bracket is still declared as an import
  • asked the oracle at most once

Ranked by hand

One reviewer put one attempt from each model in order, best first, ties left tied.

  1. #1 tie
    Gemini 3.8 Flash
    Gemini 3.8 Flash's ranked attempt
  2. #1 tie
    Haiku 5.5
    Haiku 5.5's ranked attempt
  3. #1 tie
    GPT-6 Sol
    GPT-6 Sol's ranked attempt
  4. #1 tie
    GPT-6.1 Sol
    GPT-6.1 Sol's ranked attempt
  5. #1 tie
    Sonnet 5.5
    Sonnet 5.5's ranked attempt
  6. #1 tie
    Gemini 4 Argon
    Gemini 4 Argon's ranked attempt
  7. #1 tie
    Opus 5.5
    Opus 5.5's ranked attempt

Ordering by hand decides Human rank and Human rate on the leaderboard; it never affects Overall or Strict.

Every model's attempts

Run hard-compact-2026-10-162026-10-16commit 6fa9f854spend $116.04 + $54.69 AI judging, 1 unpricedOther runs: 2026-09-23T06-56-16-hard-baseline-2026-09-23, 2026-09-10T19-44-23-baseline-2026-09, 2026-09-04T20-02-10-stage2-hard-cases, 2026-09-04T19-48-31-screen-14-fixed