Self-host LocalAI

updated Aug 2026deployed by usprices checked · Jul 2026Install verified on Ubuntu 26.04 · Jul 2026 · how we test

We earn commissions when you shop through the links below. Full disclosure →

Drop-in, self-hosted OpenAI-compatible API server for running LLM, image, and voice models on consumer CPU hardware — no GPU required, with optional NVIDIA/AMD/Intel acceleration when available.

Key facts

CategoryAI Chat Interfaces

LicenseMIT

StackGo

Min RAM4096 MB

Official imageyes

Difficulty

Sourcegithub.com/mudler/LocalAI

Websitelocalai.io

Our recommendation

Pick LocalAI when you want a drop-in OpenAI-compatible API you can run without a GPU — it serves LLM, image, and voice models on ordinary CPU hardware, so existing OpenAI-SDK code points at your server with just a URL change. It's the most API-compatible backend here (4 GB). It's a runtime first: its built-in web UI covers basic chat, but you'll usually pair it with a fuller UI or call it from your own code. If you only ever run text models and want the leanest setup, Ollama is simpler.

Follow the LocalAI deploy guide →

What you need

Any VPS with at least 4096 MB of RAM
A domain you control — most self-hosted setups need HTTPS in front of them
Under an hour if you've used Docker before

Where to host itaffiliate disclosure

Hetzner Cloudrun it on

From $23.59/mo · 2 vCPU / 4 GB / 80 GB · EU + US

Get Hetzner Cloud

Kamaterafree trial

From $4/mo · 1 vCPU / 1 GB / 20 GB · US + EU + Asia

Start free on Kamatera →

DigitalOceanalso works on

From $6/mo · 1 vCPU / 1 GB / 25 GB · US + EU + Asia

Deploy on DigitalOcean →

Paid link — we earn a commission if you shop through it.

Install

Run these commands on your server:

# LocalAI — official CPU-only container
docker run -p 8080:8080 --name local-ai -ti localai/localai:latest-cpu  #  →  http://SERVER_IP:8080

deploy guideHow to Deploy LocalAI on a VPS vps picksBest VPS for self-hostingbenchmarked quarterly · re-priced monthly

What you take on

LocalAI's OpenAI-compatibility is genuinely useful, but "runs without a GPU" comes with the usual asterisks:

non-negotiableNo GPU means slow. CPU inference works and is the selling point, but larger models will crawl — add an NVIDIA/AMD/Intel GPU when responsiveness matters, the same as every local-inference tool here.

non-negotiableIt's a backend first. LocalAI exposes an API and ships a built-in web UI with basic chat, but for a polished multi-user chat front end pair it with Open WebUI or LibreChat, or drive it from code.

non-negotiableCompatibility isn't identity. It mirrors the OpenAI API surface, but output quality depends on the open models you load — don't expect GPT-4-class answers from a small local model just because the API shape matches.