Reaction
DS V4.1F’s claimed serving-cost edge over other models
A user argues DS V4.1F costs much less to serve than other models and says its stack is better.
TLDR
A user argues that DS V4.1F is much cheaper to serve than other models and has a better stack. They say the difference is more drastic than API costs suggest because DS doesn’t want to dedicate much capacity to inference and overprices its tokens.
Combined views
1.9K
1 Source, first seen ago