Open WebUI on NAS: Setup Guide

Open WebUI is a self-hosted, ChatGPT-style interface for running large language models locally. It connects to Ollama (a local LLM runtime) and provides a polished chat interface for models like Llama 3, Mistral, Phi-3, and others. Running it on a NAS means your AI assistant is always available on your local network, requires no API key or subscription, and keeps all data local. This guide covers deploying the Open WebUI + Ollama Docker stack on a NAS, model selection for NAS hardware, and the GPU passthrough configuration that significantly improves inference speed on capable hardware.

Hardware Reality: What to Expect from NAS LLMs

NAS hardware is not designed for LLM inference. Setting realistic expectations:

• Intel Celeron N5095 (TS-464, DS423+): CPU-only inference at 2-5 tokens/second for a 3B parameter model. Usable for non-time-critical tasks. Not suitable for interactive conversation with large (7B+) models
• AMD Ryzen R1600 (DS923+, TS-473A): Slightly faster at ~5-8 tokens/second for 3B models. The integrated GPU can assist with some models
• QNAP with PCIe GPU (TS-473A + NVIDIA GPU): Adding a dedicated GPU via PCIe (NVIDIA RTX 3060/4060 in a compatible half-height format) enables GPU inference. 40-80+ tokens/second for 7B models. This is the correct hardware approach for real-time LLM use on NAS hardware

For casual, non-real-time use (asking questions and waiting 30-60 seconds for a full response), CPU-only inference on a Celeron NAS is functional. For interactive use, a GPU or dedicated inference hardware is needed.

Step 1: Deploy Ollama and Open WebUI

Create a Docker Compose file at /volume1/docker/openwebui/docker-compose.yml (Synology) or /share/docker/openwebui/docker-compose.yml (QNAP):

version: '3.8'
services:
  ollama:
    image: ollama/ollama:latest
    container_name: ollama
    volumes:
      - ./ollama:/root/.ollama
    restart: unless-stopped

  open-webui:
    image: ghcr.io/open-webui/open-webui:main
    container_name: open-webui
    depends_on:
      - ollama
    ports:
      - 3000:8080
    environment:
      - OLLAMA_BASE_URL=http://ollama:11434
    volumes:
      - ./open-webui-data:/app/backend/data
    restart: unless-stopped

Deploy with docker compose up -d. First startup downloads the Open WebUI image (~1.5GB) and starts both services. Access Open WebUI at http://[NAS-IP]:3000. Create an admin account on first access.

Step 2: Pull a Model

After Open WebUI loads, pull a model from Ollama's library. Model selection depends on your NAS RAM:

• 4GB RAM available for Ollama: Use Phi-3 Mini (3.8B parameters, ~2.3GB) or Llama 3.2 3B (~2GB). These are the smallest capable models
• 8GB RAM available: Use Llama 3.2 3B or Mistral 7B (~4.1GB). Mistral 7B is significantly more capable than 3B models
• 16GB RAM available: Use Llama 3.1 8B (~4.7GB) or Mistral 7B. More comfortable headroom

To pull a model in Open WebUI: Admin Settings → Models → pull from Ollama library. Enter the model name (e.g. llama3.2:3b) and click Pull. Model downloads from Ollama's registry. Sizes range from 2GB to 70GB+. First pull may take 20-60 minutes depending on model size and internet speed.

Alternatively, pull from the Ollama container CLI: docker exec -it ollama ollama pull llama3.2:3b

Step 3: GPU Passthrough (QNAP PCIe Models)

QNAP NAS models with PCIe slots (TS-473A, TS-673A) can host a GPU card for hardware-accelerated inference. NVIDIA consumer GPUs (RTX 3060/4060 in low-profile form factor) work with Ollama's CUDA backend.

To enable GPU passthrough in the Compose file, modify the Ollama service:

  ollama:
    image: ollama/ollama:latest
    container_name: ollama
    volumes:
      - ./ollama:/root/.ollama
    deploy:
      resources:
        reservations:
          devices:
            - driver: nvidia
              count: 1
              capabilities: [gpu]
    restart: unless-stopped

This requires the NVIDIA Container Toolkit installed on the NAS host. On QNAP, this is available as a QTS package for supported GPU models. After configuration, verify GPU usage: docker exec -it ollama ollama ps. Running models should show GPU allocation.

🇦🇺 Australian Users: Hardware Notes

Recommended hardware configurations for local LLM on NAS in Australia (March 2026):

• QNAP TS-473A (~$1,269) + NVIDIA RTX 3060 12GB ($1269-$1576 AUD): Best self-hosted LLM NAS platform in the current AU lineup. AMD Ryzen CPU, PCIe slot for GPU, 8GB RAM expandable. The RTX 3060 12GB handles 7B-13B models at 40-80 tokens/second. Total cost ~$1,700-1,800 AUD
• Intel Celeron NAS (TS-464, DS423+). CPU only: Usable for 3B models at 2-5 tokens/second. Acceptable for summarisation tasks and non-interactive queries. Not suitable for real-time conversation with capable models

If you want local LLM inference as a primary use case rather than an add-on, a dedicated mini-PC with integrated GPU (Intel Core Ultra or AMD Ryzen with strong integrated graphics) or a PC with a used NVIDIA card provides better price/performance than a NAS with GPU card.

See the best NAS for local LLM guide for a complete hardware comparison across AI workloads.

Related reading: our NAS buyer's guide and our NAS explainer.

Use our free NAS Sizing Wizard to get a personalised NAS recommendation.