weft · Kimi K3 run 1

36.1Library score
Pass rate84.8%
Simplicity49.1
Problems $$11.16
Library $$16.20
The library
weft · designed by Kimi K3 (mini-SWE)4,415 source lines
ImplementerPass rateSimplicityCostScore
DeepSeek V4.1 Flash · mini-SWE86.9%45.5$2.12
35.6
GLM 5.3 Flash · mini-SWE81.3%52.8$6.50
36.4
GPT-5.6 Luna · Codex86.2%49.1$2.54
36.4
ProblemDeepSeek V4.1 FlashGLM 5.3 FlashGPT-5.6 LunaMean
bag-rules14.926.825.622.4ninja-build19.623.331.424.8mail-addresses47.139.156.247.5property-paths45.853.637.945.8formula-ledger32.339.630.334.1record-spans54.874.740.856.8block-events41.855.149.548.8command-errors21.522.530.424.8route-tags52.635.642.943.7template-audit34.121.228.828.0grammar-inventory31.425.427.928.2grammar-trace49.840.446.045.4nesting-guard21.438.324.828.2unicode-spans31.413.337.027.2