arazu · Opus 5.5 run 3
43.2Library score
Pass rate79.5%
Simplicity68.9
Problems $$14.73
Library $$18.34
The library
arazu · designed by Opus 5.5 (mini-SWE)12,910 source lines
ImplementerPass rateSimplicityCostScore
DeepSeek V4.1 Flash · mini-SWE76.9%64.6$4.25
GLM 5.3 Flash · mini-SWE84.0%71.7$6.75
GPT-5.6 Luna · Codex77.6%70.7$3.73
ProblemDeepSeek V4.1 FlashGLM 5.3 FlashGPT-5.6 LunaMean
lander-lidar70.765.164.566.8scene-contact42.862.155.153.3world-contact43.546.50.030.0aircraft-raycast71.954.150.158.7capsule-training30.534.723.429.5physics-tick22.90.012.811.9offset-strike49.568.248.155.3layered-laser70.292.893.185.4swept-cargo40.145.028.437.8compound-scanner57.569.169.365.3sensor-transit-ledger47.576.557.960.6articulated-creature0.427.961.630.0cable-replay0.043.40.014.5damping34.743.640.239.5one-way42.841.765.249.9material-braking38.536.236.137.0platformer-ccd42.960.458.654.0grab-replay59.445.251.952.2keyed-platform-drop40.833.838.737.7ccd-wall16.711.817.015.2motor-target25.613.620.920.1rope-cut55.068.514.045.8