umami · DeepSeek V4 Pro run 2

45.7Library score
Pass rate95.7%
Simplicity51.1
Problems $$6.22
Library $$0.26
The library
umami · designed by DeepSeek V4 Pro (mini-SWE)928 source lines
ImplementerPass rateSimplicityCostScore
DeepSeek V4.1 Flash · mini-SWE98.5%54.9$2.28
53.0
GLM 5.3 Flash · mini-SWE95.8%51.5$2.58
43.9
GPT-5.6 Luna · Codex92.7%46.8$1.36
40.3
ProblemDeepSeek V4.1 FlashGLM 5.3 FlashGPT-5.6 LunaMean

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