deepseek-v4.1POST
4×RTX Pro 6000 本地跑 V4.1 Flash：单并发 250–300 tok/s
作者放出优化仓库，称 4 张 RTX Pro 6000 上单并发 250–300 tok/s、8 并发 1000+ tok/s，长上下文与视觉/engram 卸载到 SSD。
0xSero
@0xSero11 天前收录原推发布于 2026.10.18
0xSero @0xSero原推发布于 2026.10.18
Calling on 4x RTX Pro 6000 owners Optimised Ds4.1-Flash-1017 repo. I’m getting 250-300 tok/s c1 and 1000+ tok/s at c8 TONS of context and concurrency, vision and engram offloading to ssd, this does p much all my work https://github.com/sybil-solutions/deepseek-v4.1-flash-4x-rtx-pro-6000
x.com/0xsero/status/2111506865972801701采集于 2026.10.18 16:30