–
roadkill · No library run 1
8.0Library score
Pass rate51.9%
Simplicity28.0
Problems $$5.90
Implementers get no library and write everything from scratch.
ImplementerPass rateSimplicityCostScore
DeepSeek V4.1 Flash · mini-SWE56.1%27.0$1.45
GLM 5.3 Flash · mini-SWE45.0%29.1$3.41
GPT-5.6 Luna · Codex52.0%28.2$1.04
ProblemDeepSeek V4.1 FlashGLM 5.3 FlashGPT-5.6 LunaMean
transit-route-check1.50.71.81.3live-session10.20.00.03.4charging-routes2.90.00.00.9link-time-bins12.816.58.512.6explicit-fleet8.90.02.43.8braess-network13.510.013.812.4signal-plan-sweep15.10.017.510.9merge-spillback5.30.04.23.2incident-guidance2.816.72.67.4corridor-replications3.70.20.01.3speed-limit-control10.20.022.911.0reactive-signals22.116.443.427.3signal-progression7.40.09.05.5gridlock-gating23.118.615.319.0workzone-recovery1.72.12.22.0district-los3.36.86.75.6