weft · DeepSeek V4 Pro run 1

29.8Library score
Pass rate82.6%
Simplicity41.6
Problems $$9.76
Library $$0.30
The library
weft · designed by DeepSeek V4 Pro (mini-SWE)2,898 source lines
ImplementerPass rateSimplicityCostScore
DeepSeek V4.1 Flash · mini-SWE92.4%39.9$3.70
34.5
GLM 5.3 Flash · mini-SWE70.3%42.0$3.80
23.0
GPT-5.6 Luna · Codex85.2%42.9$2.26
31.9
ProblemDeepSeek V4.1 FlashGLM 5.3 FlashGPT-5.6 LunaMean
bag-rules34.520.627.727.6ninja-build22.122.825.523.5mail-addresses42.10.945.929.6property-paths62.140.046.049.4formula-ledger45.230.034.336.5record-spans39.042.130.637.2block-events37.832.639.236.5command-errors20.913.927.520.8route-tags30.921.435.329.2template-audit27.424.529.527.1grammar-inventory27.730.834.631.0grammar-trace42.10.051.231.1nesting-guard19.418.29.615.7unicode-spans31.423.69.821.6