deepseek-v4.1IMAGE
双 RTX 3090 跑 552B V4.1 Flash：34.9 tok/s
用 lxe 的 dsx 把热专家放显存、冷 T16 专家放内存，双 3090+128GB 内存解码 34.9 tok/s，需 CUDA 13 和约 305GB NVMe。
Sympathy
@SympathyLabs9 天前收录原推发布于 2026.10.23
Sympathy @SympathyLabs原推发布于 2026.10.23
running 552b deepseek-v4.1-flash-1017 on two rtx 3090s + 128gb ram? @lxe's dsx keeps hot experts in vram and cold t16 experts in ram for 34.9 tok/s decode (36.8 with --defer-rank 3). mit-licensed; needs cuda 13 and ~305gb nvme. https://github.com/lxe/dsx
x.com/sympathylabs/status/2113377753286468048采集于 2026.10.23 12:08