deepseek-v4.1POST

Maya and Strata bring DeepSeek V4.1 Flash to consumer PCs via CPU offloading

Two CPU-offloading projects support DeepSeek V4.1 Flash 1017; the author tested Qwen3.8 on one 3090 with 64GB DDR4 at 100 tok/s decode.

米柒说｜Local AI & Agents

@meseven_journeyArchived 2 days agoOriginally posted 2026.10.27

米柒说｜Local AI & Agents @meseven_journeyOriginally posted 2026.10.27

Open original on X

让开源超大模型能部署到普通PC上的2个项目🔥🔥🔥 充分利用VRAM + RAM + SSD ，也就是 CPU offloading Project Maya 主要支持 > DeepSeek-V4.1-Flash-1017 > GLM-5.3-Flash > MiniMax-M3.1-Flash(最新支持) Strata 主要支持 > DeepSeekV4.1-Flash-1017 > Qwen3.8-Flash-Next > Qwen3.6-35B-A3B 我的实测🔥: 单3090 / 64GB DDR4 / qwen3.8-flash-next / strata: decode 100 toks、prefill 1400 toks

This post quotes another post: “Maya v1.1.29 adds local MiniMax M3.1 Flash support”

x.com/meseven_journey/status/2114884326102073552Archived 2026.10.27 12:08