Categories / LLM Infrastructure

DeepSeek-V4.1-Flash-1017 (552B MoE + 196B Engram) on 2x RTX 3090 + 128 GB RAM. Rust + CUDA, T16 quantization. 40+ tok/s measured with expert deferral + FP8 projections.

Last push 2 weeks ago

Created

Oct 2026

Total stars

108

Last 7 days

+8

Last 30 days

+81

Last 12 months

+108

README badge

Rusty Stars badge for lxe/dsx
[![Rusty Stars](https://rustystars.dev/badge/lxe/dsx.svg)](https://rustystars.dev/lxe/dsx)