Show HN: NixOS-DGX-Spark – Nix and NixOS on the DGX Spark | Hacker News
TL;DR AI
2 min readKey summary
A Show HN post introduced NixOS-DGX-Spark, a project that brings Nix and NixOS to Nvidia’s DGX Spark for local AI inference workloads.
It reported high token throughput using DeepSeek-based models with custom vLLM images, along with FP8-related tuning.
The discussion compared models for coding and review tasks, mentioning Hermes, Qwen 3.6 27B, and DeepSeek-V4-Flash-0731.
The example shows a practical, reproducible way to deploy and tune LLMs on specialized Nvidia hardware with Nix-based systems.