arazu · GLM 5.3 run 1

34.1Library score
Pass rate72.9%
Simplicity61.5
Problems $$17.33
Library $$37.23
The library
arazu · designed by GLM 5.3 (mini-SWE)10,240 source lines
ImplementerPass rateSimplicityCostScore
DeepSeek V4.1 Flash · mini-SWE72.3%59.5$2.86
34.0
GLM 5.3 Flash · mini-SWE75.6%64.5$10.44
38.2
GPT-5.6 Luna · Codex70.7%60.5$4.03
30.2
ProblemDeepSeek V4.1 FlashGLM 5.3 FlashGPT-5.6 LunaMean
lander-lidar39.827.93.823.8scene-contact18.216.817.517.5world-contact67.377.40.048.2aircraft-raycast26.530.124.827.1capsule-training4.55.22.84.2physics-tick8.717.022.216.0offset-strike51.467.345.054.6layered-laser33.623.222.926.6swept-cargo3.232.010.215.1compound-scanner62.364.457.661.4sensor-transit-ledger45.583.165.764.8articulated-creature0.93.30.71.6cable-replay26.428.634.629.9damping43.440.525.136.3one-way44.332.254.743.7material-braking29.142.430.834.1platformer-ccd36.737.337.137.0grab-replay40.928.549.339.6keyed-platform-drop44.832.031.736.1ccd-wall25.949.041.038.6motor-target36.429.027.430.9rope-cut59.372.159.163.5