weft · DeepSeek V4 Pro run 3

30.3Library score
Pass rate85.3%
Simplicity44.3
Problems $$8.80
Library $$0.34
The library
weft · designed by DeepSeek V4 Pro (mini-SWE)2,429 source lines
ImplementerPass rateSimplicityCostScore
DeepSeek V4.1 Flash · mini-SWE89.1%40.4$3.17
32.6
GLM 5.3 Flash · mini-SWE83.6%44.9$3.74
24.4
GPT-5.6 Luna · Codex82.8%47.7$1.89
33.9
ProblemDeepSeek V4.1 FlashGLM 5.3 FlashGPT-5.6 LunaMean
bag-rules31.732.729.431.3ninja-build35.422.332.130.0mail-addresses33.20.053.829.0property-paths45.233.254.344.2formula-ledger40.546.540.242.4record-spans35.234.936.735.6block-events45.043.547.545.4command-errors26.025.524.525.3route-tags25.027.529.927.5template-audit20.40.030.717.0grammar-inventory24.527.024.825.4grammar-trace47.40.040.829.4nesting-guard18.922.06.715.9unicode-spans28.526.523.326.1