12GB 显存 + 80GB 内存跑 V4.1 Flash：7.1 tok/s

KASperience❄️🌍 @KASperiencexyz原推发布于 2026.10.25

在 X 上打开原推

first successful output today, i recommend low reasoning max (this was like 100 min run... with 7.1 tok/s): same prompt. same machine (12gb vram, 80gb ram, but 64 is ok for dedicated box). new model: deepseek v4.1 flash via strata #strata #llm #deepseek re-upload since there was some stuttering from overloaded gpu

这条是引用着另一条写的：x.com/kasperiencexyz/status/2108161567661453562