weft · GPT-6 Astra run 1

35.9Library score
Pass rate80.8%
Simplicity53.2
Problems $$9.58
Library $$3.08
The library
weft · designed by GPT-6 Astra (mini-SWE)1,833 source lines
ImplementerPass rateSimplicityCostScore
DeepSeek V4.1 Flash · mini-SWE88.1%50.0$2.48
39.8
GLM 5.3 Flash · mini-SWE77.0%55.0$4.78
32.7
GPT-5.6 Luna · Codex76.9%54.6$2.32
35.2
ProblemDeepSeek V4.1 FlashGLM 5.3 FlashGPT-5.6 LunaMean
bag-rules36.918.732.029.2ninja-build24.639.933.432.6mail-addresses37.550.41.229.7property-paths51.623.943.139.6formula-ledger33.346.332.637.4record-spans51.758.369.159.7block-events60.658.162.560.4command-errors15.10.026.413.8route-tags54.925.867.749.5template-audit40.243.840.941.7grammar-inventory34.839.125.033.0grammar-trace61.50.14.121.9nesting-guard21.123.528.324.3unicode-spans34.129.427.230.2