# NVIDIA DGX Spark setup for local AI

Reviewed by the Contractor AI Shop research team · Last updated Oct 15, 2026  
HTML: https://contractoraishop.com/guides/dgx-spark-setup

The [DGX Spark](https://www.nvidia.com/en-us/products/workstations/dgx-spark/) is a desktop Grace Blackwell system with 128GB of unified memory, so it can serve models to the office.

## 1. First boot

Follow the official [DGX Spark User Guide](https://docs.nvidia.com/dgx/dgx-spark/) for setup, networking and updates.

## 2. Use NVIDIA playbooks

[build.nvidia.com/spark](https://build.nvidia.com/spark) has step-by-step playbooks (Ollama, Open WebUI, vLLM, fine-tuning).

## 3. Serve models

Start simply with [Ollama](https://docs.ollama.com). Use [vLLM](https://docs.vllm.ai) for higher-throughput, multi-user, OpenAI-compatible serving.

## 4. Team access

Put [Open WebUI](https://docs.openwebui.com/getting-started/quick-start/) in front with user accounts. Keep it on the LAN or VPN, not open to the internet.

Independent guide by Contractor AI Shop. Not affiliated with listed vendors. Follow the linked official docs for current steps.