deepseek-v4.1POST
DeepSeek V4.1 Flash fully local on 2 DGX Sparks: 50-66 tok/s per request
Fully local V4.1 Flash setup with two DGX Sparks plus an x86 box hits 50-66 tok/s per request, ~105-119 combined, 2,200 tok/s prefill and 2.5M-token KV.
Nick Mermiges, Esq. 🦁🏈
@NicoTheGrecoArchived 8 days agoOriginally posted 2026.10.24
Nick Mermiges, Esq. 🦁🏈 @NicoTheGrecoOriginally posted 2026.10.24
DeepSeek V4.1 Flash, fully local: 2 DGX Sparks + an x86 box. Drafter on 2 NVLinked 3090s, Engram in its RAM via RDMA. ~50-66 tok/s per request, ~105-119 combined, 2,200 tok/s prefill, 2.5M-token KV. https://github.com/ProprietaryLegal/deepseek-v41-flash-2x-dgx-spark-progress
x.com/nicothegreco/status/2113774487220949433Archived 2026.10.24 12:07