deepseek-v4.1IMAGE
DeepSeek V4.1 Flash in llama.cpp with Unsloth PR hits 31.6 tok/s
Hands-on test of DeepSeek V4.1 Flash in llama.cpp using the Unsloth PR reached 31.6 tok/s, which the poster calls quite fast.
きしだൠ(K1S)
@kisArchived 10 days agoOriginally posted 2026.10.22

きしだൠ(K1S) @kisOriginally posted 2026.10.22
Unsloth版PRを使うと、31.6tok/secでた。かなり速い。 llama.cppでDeepSeek V4.1 FlashならUnsloth版PRだな。
This post quotes another post: x.com/kis/status/2111564981905072551
x.com/kis/status/2113046824814833901Archived 2026.10.22 11:58