–
plaid · No library run 1
37.4Library score
Pass rate93.6%
Simplicity42.6
Problems $$8.55
Implementers get no library and write everything from scratch.
ImplementerPass rateSimplicityCostScore
DeepSeek V4.1 Flash · mini-SWE98.4%42.3$1.54
GLM 5.3 Flash · mini-SWE97.0%44.4$4.56
GPT-5.6 Luna · Codex85.6%41.3$2.45
ProblemDeepSeek V4.1 FlashGLM 5.3 FlashGPT-5.6 LunaMean
ledger-balance37.039.333.636.6typed-settings46.844.841.644.4keymap-resolver36.439.039.438.3template-trace49.648.941.346.6recipe-graph54.454.419.842.9scoped-eval44.80.030.625.1temporal-normalizer35.333.132.933.7script-recovery31.236.734.534.2stack-audit33.435.135.234.6query-resolution55.444.952.550.9selection-flattener37.939.537.438.3schema-resolver39.348.035.741.0grammar-crossref39.343.539.740.8config-digest33.135.916.428.5shape-atlas37.638.80.125.5