weft · Kimi K3 run 3

36.1Library score
Pass rate83.7%
Simplicity49.4
Problems $$11.26
Library $$39.05
The library
weft · designed by Kimi K3 (mini-SWE)5,908 source lines
ImplementerPass rateSimplicityCostScore
DeepSeek V4.1 Flash · mini-SWE89.5%43.6$2.42
35.9
GLM 5.3 Flash · mini-SWE78.4%53.7$6.32
36.5
GPT-5.6 Luna · Codex83.3%50.9$2.52
36.1
ProblemDeepSeek V4.1 FlashGLM 5.3 FlashGPT-5.6 LunaMean
bag-rules32.817.935.728.8ninja-build26.631.336.131.3mail-addresses40.546.446.044.3property-paths53.436.878.756.3formula-ledger32.440.332.535.1record-spans39.764.036.646.7block-events55.962.247.955.3command-errors15.028.223.322.2route-tags56.234.431.040.5template-audit31.235.925.931.0grammar-inventory19.730.028.726.1grammar-trace43.351.934.143.1nesting-guard20.90.415.812.4unicode-spans34.930.932.432.7