Best home AI server setup
Last updated
The best home AI server setup is one box on wired Ethernet with Ollama for models, Open WebUI as the family chat page, a VPN for access from outside and automatic backups. Olares OS ships these; on a DIY server you install them.
Olares One → Home AI server → Ollama · Open WebUI · Immich → Family AI that stays at home
What goes into the best home AI server setup?
Every home setup has the same parts, whatever the hardware. You need a model runner, usually Ollama, a chat page such as Open WebUI, one login per person, a VPN for access away from home and automatic backups. Put the box on a ventilated shelf with wired Ethernet, which is steadier than Wi-Fi for a machine that never turns off.
How do you set up Olares One?
Olares One sets up from a phone and a browser. According to Olares' FAQ, activation needs an internet connection, an Olares ID and the LarePass app. After that you install apps from the Olares Market and add family accounts. Remote access works without a static IP, through Olares' built-in VPN or its reverse proxy.
- Plug in power and Ethernet.
- Activate the box with LarePass and an Olares ID.
- Install Ollama and Open WebUI from the Olares Market. Both are one-click installs.
- Download a first model sized for 24 GB of VRAM.
- Create accounts. Each person connects with their own LarePass app.
- Turn on remote access. Olares includes a VPN and supports Tailscale.
- Schedule encrypted backups to an external location, or SMB backup to a NAS.
After activation the box can run entirely offline on your home network. Installing new apps needs internet.
How does a DIY setup differ?
A DIY setup needs the same pieces, installed by hand. You install a Linux distribution such as Ubuntu, then the NVIDIA driver and Ollama. Open WebUI usually runs in Docker and keeps its own user accounts. Tailscale or another VPN handles remote access, and a backup tool you choose handles copies. Test a restore before you trust it.
The software costs nothing. The upkeep is yours, from driver updates to container updates. The build or buy guide weighs that cost.
Which models should a home server start with?
Start with a small fast model for quick tasks, then add one larger model for quality. An 8B model needs about 5 GB and a 32B model about 19-20 GB (approximate rule of thumb at 4-bit). We'd make the larger one a mixture-of-experts model, because only part of it is active per token, which keeps answers quick on 24 GB.
In Olares' own test, Qwen3-30B-A3B ran at 157 tokens per second for a single request on Olares One. Then add photo search with Immich, covered in the NAS guide, and per-person logins, covered in family accounts.
Questions & answers
Do I need to open ports on my router?
No. A VPN such as Tailscale lets your phone reach the server from anywhere without port forwarding. According to Olares' FAQ, Olares offers a VPN and a reverse proxy and does not need a static IP.
Which chat apps run on Olares One?
Open WebUI installs from the Olares Market, and Olares' FAQ also names LobeHub. Desktop apps such as LM Studio stay on your own computer and connect to the server's API.
Sources for this entry
- Olares One FAQ (Olares docs) (Company-published)
- Olares One product page (Company-published)
- Olares blog: local AI hardware benchmarking (Oct 2025) (Company-published)
- Olares One: maker's page (Maker-published)
Continue this reading path
How the related websites connect
Cite this entry
Best Home AI Server. “Best home AI server setup.” Content reviewed 13 November 2026.
https://besthomeaiserver.com/best-home-ai-server-setup/