deepseek-v4.1POST
User reports DeepSeek V4.1 Flash thinks too long before acting
Hands-on complaint: even at 150 tok/s, the model spends too long thinking before doing anything, which hurts a Flash-tier model.
Elshayib
@elshayib_Archived 11 days agoOriginally posted 2026.10.17
Elshayib @elshayib_Originally posted 2026.10.17
DeepSeek v4.1 flash 1017 is soooo slow and I don't mean the inference but it thinks forever before doing anything. As a flash model this is extremely bad it's supposed to be fast and even if it runs at 150 tok/s that won't matter if it thinks all day.
x.com/elshayib_/status/2111480830010814700Archived 2026.10.18 16:43