deepseek-v4.1IMAGE
DeepSeek V4.1 Flash-1017 hits ~70 tok/s on three V100 32GB GPUs
Local deployment report: three V100 32GB cards plus DDR4 hit ~70 tok/s decode and ~446 tok/s prefill at 256k context using the new Maya-XS quant.
AJ
@ItsmeAjayKVArchived 9 days agoOriginally posted 2026.10.23
AJ @ItsmeAjayKVOriginally posted 2026.10.23
Wow, this is DeepSeek-v4.1-flash-1017 running on 3 Tesla V100 32GB and DDR4 ram. ~ 70 tok/s decode ~ 446 tok/s prefill - 256k token window Project Maya has now improved a lot, @PeasantSmith even created a new quant Maya-XS to make this happen. I'm downloading weights now, will be testing how this will perform on my 24GB 3090 + 64GB RAM.
This post quotes another post: “763B DeepSeek V4.1-Flash-1017 running on 3x Tesla V100 32GB”
x.com/itsmeajaykv/status/2113400773803741293Archived 2026.10.23 12:05