P

arazu · Production library run 3

41.3Library score
Pass rate76.7%
Simplicity67.4
Problems $$25.04

Implementers use the real production library (rapier).

ImplementerPass rateSimplicityCostScore
DeepSeek V4.1 Flash · mini-SWE78.4%60.8$3.73
39.2
GLM 5.3 Flash · mini-SWE79.5%70.9$15.37
44.3
GPT-5.6 Luna · Codex72.1%70.6$5.95
40.2
ProblemDeepSeek V4.1 FlashGLM 5.3 FlashGPT-5.6 LunaMean
lander-lidar56.150.948.952.0scene-contact33.138.146.839.3world-contact27.442.231.833.8aircraft-raycast44.963.554.954.4capsule-training28.924.46.219.8physics-tick20.112.018.416.8offset-strike67.477.960.968.7layered-laser78.871.078.176.0swept-cargo38.043.438.239.9compound-scanner36.857.545.946.8sensor-transit-ledger38.623.368.343.4articulated-creature39.631.90.724.0cable-replay0.045.20.015.1damping41.161.042.248.1one-way30.327.252.436.6material-braking33.446.833.437.9platformer-ccd34.260.451.848.8grab-replay62.252.658.557.8keyed-platform-drop57.041.334.644.3ccd-wall15.17.34.99.1motor-target6.015.618.513.4rope-cut73.281.589.781.5