heretical · DeepSeek V4 Pro run 3

28.5Library score
Pass rate87.2%
Simplicity37.5
Problems $$10.30
Library $$0.37
The library
heretical · designed by DeepSeek V4 Pro (mini-SWE)1,660 source lines
ImplementerPass rateSimplicityCostScore
DeepSeek V4.1 Flash · mini-SWE89.5%38.2$3.71
30.6
GLM 5.3 Flash · mini-SWE82.2%39.4$3.89
26.2
GPT-5.6 Luna · Codex89.7%35.0$2.70
28.7
ProblemDeepSeek V4.1 FlashGLM 5.3 FlashGPT-5.6 LunaMean

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