roadkill · Kimi K3 run 2

11.9Library score
Pass rate60.1%
Simplicity35.2
Problems $$8.27
Library $$3.64
The library
roadkill · designed by Kimi K3 (mini-SWE)2,269 source lines
ImplementerPass rateSimplicityCostScore
DeepSeek V4.1 Flash · mini-SWE64.1%37.8$2.13
14.7
GLM 5.3 Flash · mini-SWE58.5%32.1$4.45
8.7
GPT-5.6 Luna · Codex57.8%35.2$1.70
12.5
ProblemDeepSeek V4.1 FlashGLM 5.3 FlashGPT-5.6 LunaMean
transit-route-check1.66.85.14.5live-session24.90.032.919.3charging-routes41.10.81.614.5link-time-bins25.922.822.423.7explicit-fleet5.38.00.14.5braess-network7.311.416.511.7signal-plan-sweep3.20.016.26.5merge-spillback4.43.63.53.9incident-guidance12.08.810.010.3corridor-replications21.41.92.28.5speed-limit-control19.30.014.411.2reactive-signals31.028.326.228.5signal-progression4.08.814.29.0gridlock-gating17.018.318.317.9workzone-recovery0.010.02.14.0district-los16.29.914.113.4