canon · Kimi K3 run 2
56.8Library score
Pass rate91.6%
Simplicity67.6
Problems $$5.06
Library $$7.39
The library
canon · designed by Kimi K3 (mini-SWE)3,954 source lines
ImplementerPass rateSimplicityCostScore
DeepSeek V4.1 Flash · mini-SWE93.9%67.4$1.87
GLM 5.3 Flash · mini-SWE92.6%73.9$1.93
GPT-5.6 Luna · Codex88.3%61.5$1.25
ProblemDeepSeek V4.1 FlashGLM 5.3 FlashGPT-5.6 LunaMean
profile-patch100.087.528.972.1thermostat-codec50.172.937.553.5tagged-config79.777.065.073.9rtl-pin-lock65.670.195.076.9pipeline-provenance38.649.330.239.4boundary-views48.234.942.541.9ml-run-config82.889.775.082.5ha-triggers78.678.362.073.0env-config68.195.065.676.2paper-poster45.757.745.449.6typed-client48.346.845.246.8cli-dispatch42.546.539.742.9trading-events20.982.17.236.7llm-contracts56.74.328.829.9