deepseek-v4.1POST
Daily-driver test: GLM 5.3 Flash hits 177.2 tok/s vs DeepSeek V4.1 Flash at 140.9 tok/s
After using DeepSeek V4.1 Flash as a daily model for 19.5 hours and switching back to GLM 5.3 Flash, the same 4 boxes and 8 chats ran at 177.2 tok/s for GLM versus 140.9 tok/s for DeepSeek, though DeepSeek's builds looked nicer.
CryptoBandit
@CryptoBandit100Archived 10 days agoOriginally posted 2026.10.22
CryptoBandit @CryptoBandit100Originally posted 2026.10.22
Ran DeepSeek V4.1 Flash as my daily model for 19.5 hours, then went back to GLM 5.3 Flash. Same 4 boxes, 8 chats: GLM 177.2 tok/s DeepSeek 140.9 tok/s GLM is faster. DeepSeek's builds look nicer. Which one would you run every day, and why? 👇
x.com/cryptobandit100/status/2113057753409138917Archived 2026.10.22 11:58