deepseek-v4.1IMAGE

DeepSeek V4.1 Flash at 2.25bpw on an RTX 5060 Ti 16GB: 7.6 tok/s, 77-minute pagoda task

A 2.25bpw quant of DeepSeek V4.1 Flash on an OCuLink RTX 5060 Ti 16GB reaches 7.6 tok/s and finishes a pagoda task in 77 minutes, matching quality but about 4x slower than the comparison model.

yutacar@niigata

@yutacarArchived 1 day agoOriginally posted 2026.10.26

DeepSeek V4.1 Flash at 2.25bpw on an RTX 5060 Ti 16GB: 7.6 tok/s, 77-minute pagoda task

yutacar@niigata @yutacarOriginally posted 2026.10.26

Open original on X

ミニPC＋OCuLink RTX 5060 Ti 16GBで、ローカルLLM 2モデルを比較しました UD-IQ4_XS：30tok/s、五重塔70分 DeepSeek V4.1 Flash 2.25bpw：7.6tok/s、五重塔77分品質は互角、速さはUDが約4倍。DeepSeekも0.2.5＋SSD追加で完走できるように #ローカルLLM #Strata

x.com/yutacar/status/2114676570539667523Archived 2026.10.27 12:10