–
roadkill · No library run 2
7.8Library score
Pass rate57.9%
Simplicity25.2
Problems $$4.79
Implementers get no library and write everything from scratch.
ImplementerPass rateSimplicityCostScore
DeepSeek V4.1 Flash · mini-SWE60.7%20.5$1.49
GLM 5.3 Flash · mini-SWE58.0%25.5$2.22
GPT-5.6 Luna · Codex54.9%30.0$1.09
ProblemDeepSeek V4.1 FlashGLM 5.3 FlashGPT-5.6 LunaMean
transit-route-check2.80.01.81.5live-session5.10.06.53.9charging-routes1.70.00.00.6link-time-bins2.10.014.45.5explicit-fleet11.40.013.08.1braess-network12.40.015.59.3signal-plan-sweep10.30.010.26.8merge-spillback7.22.610.56.8incident-guidance2.516.10.06.2corridor-replications2.53.10.01.9speed-limit-control5.80.06.84.2reactive-signals14.317.946.826.3signal-progression6.66.38.67.2gridlock-gating24.119.420.921.5workzone-recovery2.02.43.82.7district-los13.510.112.712.1