plaid · Kimi K3 run 3

36.2Library score
Pass rate92.8%
Simplicity49.5
Problems $$12.16
Library $$5.14
The library
plaid · designed by Kimi K3 (mini-SWE)1,879 source lines
ImplementerPass rateSimplicityCostScore
DeepSeek V4.1 Flash · mini-SWE99.0%47.0$2.65
46.1
GLM 5.3 Flash · mini-SWE91.1%50.2$7.17
22.4
GPT-5.6 Luna · Codex87.4%51.6$2.34
40.0
ProblemDeepSeek V4.1 FlashGLM 5.3 FlashGPT-5.6 LunaMean
ledger-balance49.515.647.437.5typed-settings46.30.027.724.7keymap-resolver45.50.043.829.7template-trace37.942.233.337.8recipe-graph46.20.019.421.9scoped-eval45.70.00.015.3temporal-normalizer45.339.237.240.5script-recovery50.80.044.631.8stack-audit44.346.742.844.6query-resolution52.148.466.955.8selection-flattener56.90.053.636.8schema-resolver44.645.250.046.6grammar-crossref46.947.639.144.5config-digest37.60.046.628.1shape-atlas42.151.847.747.2