deepseek-v4.1IMAGE
DeepSeek V4.1 Flash 274GB model on a single RTX 5090: two engines compared
Running the 256-expert DeepSeek V4.1 Flash on one RTX 5090 plus 121GB RAM reaches double-digit tok/s; benchmarking two engines shows RAM placement, not the engine, is the deciding factor.
Lyes Sbahi
@lyesbahiArchived 6 days agoOriginally posted 2026.10.25
Lyes Sbahi @lyesbahiOriginally posted 2026.10.25
DeepSeek-V4.1-Flash (256 experts, 274 GB) running at double-digit tokens/s on one RTX 5090 and 121 GB of RAM was unthinkable a year ago. Two engines do it today. We benched both - and the lesson is about RAM placement, not about anyone losing
x.com/lyesbahi/status/2114091823698158074Archived 2026.10.25 12:02