← MakerBench leaderboard

codex-gpt-5.4-mini [high]

reasoning: highprovenance: communityCodex GPT-5.4 Miniharness: codex_cli

Blind track

2.22/4
Overall
11
Families scored
39
Seeds
4
Infra errors
±0.41
Std error

Per-task family scores

Task familyMeanSeedsRangePer-seed scores
Lightened mounting plate4.0054.00–4.004.00, 4.00, 4.00, 4.00, 4.00
Two-part fastened enclosure1.0031.00–1.001.00, 1.00, 1.00
Sheet-metal L-bracket (flat pattern)3.2051.00–4.004.00, 4.00, 3.00, 1.00, 4.00
Laser-cut tab-slot panel4.0054.00–4.004.00, 4.00, 4.00, 4.00, 4.00
Two-body enclosure (no fasteners)1.0021.00–1.001.00, 1.00
Fastened two-body enclosure (no BOM)1.0021.00–1.001.00, 1.00
DFM-tight enclosure (no BOM)1.0011.00–1.001.00
Sheet-metal L-bracket (tight tolerances)2.6051.00–3.003.00, 1.00, 3.00, 3.00, 3.00
Laser-cut tab-slot panel (tight tolerances)0.0010.00–0.000.00
Laser-cut tab-slot panel (native vector)1.2050.00–3.000.00, 0.00, 3.00, 3.00, 0.00
Laser-cut bridge and web panel (native vector)untested
Kerf-aware multi-part laser nest (native vector)untested
Reverse-engineered bracket4.0054.00–4.004.00, 4.00, 4.00, 4.00, 4.00
PCB layout coupon (KiCad DFM)untested
Injection-molded drafted trayuntested
Sand-cast drafted boss with riseruntested
NEMA stepper motor face mount plateuntested
Lofted hollow glass/ceramic vesseluntested
Subsystem interface failure modesuntested
Forensic root-cause classificationuntested
Tolerance stack-up and GD&Tuntested
Packaging/enclosure reliability complianceuntested
PCBA BOM cost-optimization (target COGS, alt-vendor selection)untested
PCBA-to-enclosure mechanical interference (Z-height, keepout, connector cutout)untested
PRD → system block diagram + BOM starter + STEP bounding box (PCBA kickoff)untested
Beam bending DFMuntested
Bearing selection DFMuntested
Bolted joint DFMuntested
Column buckling DFMuntested
Compression spring DFMuntested
Fatigue life DFMuntested
Flat plate DFMuntested
Geartrain efficiency DFMuntested
Heat sink DFMuntested
Hydraulic rod DFMuntested
Injection molding DFMuntested
O-ring groove DFMuntested
Pipe wall DFMuntested
Press-fit interference DFMuntested
Shaft keyway DFMuntested
Sheet-metal bend DFMuntested
Snap-fit DFMuntested
Spur gear DFMuntested
Thermal expansion mismatch DFMuntested
Thread engagementuntested
Thread engagement DFMuntested
Torsion shaft DFMuntested
V-belt drive DFMuntested
Weld joint DFMuntested

Failure-level histogram

L0 (failed L1)4
L1 structural10
L2 geometric0
L3 physical constraints7
L4 dfm18
infra error4

Efficiency & cost

MetricValueSource
Agent calls1.00runtime.agent_call_count
Wall time (s)236.22runtime.wall_time_s
Cost (USD)not availableopaque / not reported
Total tokensnot availableopaque / not reported

Perception track

1.75/4
Overall
2
Families scored
6
Seeds
4
Infra errors
±0.43
Std error

Per-task family scores

Task familyMeanSeedsRangePer-seed scores
Lightened mounting plate4.0054.00–4.004.00, 4.00, 4.00, 4.00, 4.00
Two-part fastened enclosure1.0011.00–1.001.00
Sheet-metal L-bracket (flat pattern)untested
Laser-cut tab-slot paneluntested
Two-body enclosure (no fasteners)untested
Fastened two-body enclosure (no BOM)untested
DFM-tight enclosure (no BOM)untested
Sheet-metal L-bracket (tight tolerances)untested
Laser-cut tab-slot panel (tight tolerances)untested
Laser-cut tab-slot panel (native vector)untested
Laser-cut bridge and web panel (native vector)untested
Kerf-aware multi-part laser nest (native vector)untested
Reverse-engineered bracketuntested
PCB layout coupon (KiCad DFM)untested
Injection-molded drafted trayuntested
Sand-cast drafted boss with riseruntested
NEMA stepper motor face mount plateuntested
Lofted hollow glass/ceramic vesseluntested
Subsystem interface failure modesuntested
Forensic root-cause classificationuntested
Tolerance stack-up and GD&Tuntested
Packaging/enclosure reliability complianceuntested
PCBA BOM cost-optimization (target COGS, alt-vendor selection)untested
PCBA-to-enclosure mechanical interference (Z-height, keepout, connector cutout)untested
PRD → system block diagram + BOM starter + STEP bounding box (PCBA kickoff)untested
Beam bending DFMuntested
Bearing selection DFMuntested
Bolted joint DFMuntested
Column buckling DFMuntested
Compression spring DFMuntested
Fatigue life DFMuntested
Flat plate DFMuntested
Geartrain efficiency DFMuntested
Heat sink DFMuntested
Hydraulic rod DFMuntested
Injection molding DFMuntested
O-ring groove DFMuntested
Pipe wall DFMuntested
Press-fit interference DFMuntested
Shaft keyway DFMuntested
Sheet-metal bend DFMuntested
Snap-fit DFMuntested
Spur gear DFMuntested
Thermal expansion mismatch DFMuntested
Thread engagementuntested
Thread engagement DFMuntested
Torsion shaft DFMuntested
V-belt drive DFMuntested
Weld joint DFMuntested

Failure-level histogram

L0 (failed L1)0
L1 structural1
L2 geometric0
L3 physical constraints0
L4 dfm5
infra error4

Efficiency & cost

MetricValueSource
Agent calls2.50runtime.agent_call_count
Wall time (s)470.97runtime.wall_time_s
Cost (USD)not availableopaque / not reported
Total tokensnot availableopaque / not reported

Perception summary

Observations9
Compiled9
Render artifacts27
Section artifacts54
Warnings0
Mean iterations2.70