← MakerBench task families

Sheet-metal L-bracket (tight tolerances)

sheet_metalpack: sheet-metaltier 3blindperception

L-bracket graded for bend-allowance precision, constant gauge, developed volume, and usable flange. Promoted from the intermediate calibrators (#4/#5); reuses the sheet_metal_bracket private oracle at tighter tolerances.

Capability axes

dfm_manufacturabilitysheet_metal

Graded categories

structuralgeometricphysicsdfm

Grading ladder

1
Structural
Compiles to a single, watertight, non-empty solid.
2
Geometric
Outer dimensions / bounding box match the brief within tolerance.
3
Physical constraints
Mass, fit, and physical-constraint targets are met.
4
DFM
Manufacturable for the process (e.g. minimum wall thickness).

Full task brief (task.md) →

Per-model scores

Modelblindperception
codex-gpt-5.4-mini [xhigh]3.00—
claude-code-sonnet-54.004.00
antigravity-gemini-3.5-flash4.004.00
grok-4.54.003.00
claude-code-fable-5 [high]4.004.00
codex-gpt-5.6-terra1.004.00
claude-code-opus-4.8 [high]4.004.00
codex-gpt-5.6-sol3.003.00
codex-gpt-5.4 [xhigh]3.50—
gemini-3-flash-preview [default_or_unset]4.004.00
gemini-3.5-flash4.003.80
codex-gpt-5.4 [medium]4.003.00
codex-gpt-5.6-luna3.003.00
codex-gpt-5.4 [high]3.673.33
codex-gpt-5.4 [low]3.003.33
codex-gpt-5.5 [low]1.802.20
kimi-k2.6 [default_or_unset]4.00infra
claude-code-sonnet-4.6 [high]4.004.00
codex-gpt-5.4-mini [high]2.60—
codex-gpt-5.5 [medium]1.40—
codex-gpt-5.3-codex-spark [xhigh]1.00—
qwen3.6-max-preview [default_or_unset]1.671.67
codex-gpt-5.5 [high]1.401.80
codex-gpt-5.5 [xhigh]0.80—
codex-gpt-5.4-mini [medium]1.00—
deepseek-v4-flash [default_or_unset]2.672.67
codex-gpt-5.3-codex-spark [high]2.00—
codex-gpt-5.3-codex-spark [medium]1.00—
deepseek-v4-pro [default_or_unset]2.002.67
codex-gpt-5.4-mini [low]1.00—
grok-4.3 [default_or_unset]1.001.67
claude-code-haiku-4.5 [high]1.001.00
codex-gpt-5.3-codex-spark [low]1.00—
qwen3-max [default_or_unset]1.001.00
antigravity-gemini-3-flash0.000.00
antigravity-gemini-3.1-pro0.000.00