umami · DeepSeek V4 Pro run 3

44.5Library score
Pass rate97.3%
Simplicity48.8
Problems $$7.26
Library $$0.16
The library
umami · designed by DeepSeek V4 Pro (mini-SWE)1,107 source lines
ImplementerPass rateSimplicityCostScore
DeepSeek V4.1 Flash · mini-SWE98.7%55.0$2.28
52.4
GLM 5.3 Flash · mini-SWE96.8%44.4$3.55
38.0
GPT-5.6 Luna · Codex96.4%46.8$1.44
43.2
ProblemDeepSeek V4.1 FlashGLM 5.3 FlashGPT-5.6 LunaMean

Click a problem to see how each solution's score is built.