LocalAI
mudler/LocalAIReviewed by AI
Checked by RepoPicks Radar on
What is LocalAI?
LocalAI is a server you run yourself that loads AI models and answers requests in the same shape as the OpenAI, Anthropic and ElevenLabs online services, so programs written for those can point at your computer instead. It covers chat, transcription, speech, images, video, embeddings and object detection, works on NVIDIA, AMD, Intel or Apple chips and on the processor alone, and has a web interface, saved chats with agents and tools, user accounts and a cluster mode. There is a macOS installer; most people start it with Docker.
LocalAI is free and open source (MIT) and can replace ChatGPT and OpenRouter.
How easy is it to install LocalAI?
C · install 1 or 2 tools · macOS
- macOS C
- Linux C
The easiest route is the macOS LocalAI.dmg from the release page, but the README warns "The DMG is not signed by Apple" and tells you to run `sudo xattr -d com.apple.quarantine /Applications/LocalAI.app` before it opens, and you then pick and download a model. On Linux there is a launcher AppImage and a plain binary, while the README's own quickstart is a `docker run` command with a different image for each kind of graphics card.
LocalAI on GitHub
- Stars
- 50,296on Nov 10, 2026
- Downloads
- 223,6962,728 this week
- Growth
- +115stars in 7 days
Stars over time
Facts
- Runs on
- macOS, Linux
- You need to install
- Nothing extra
- To build from source
- Docker, NVIDIA graphics card, Go
- Disk space
- not determined
- Accounts or keys
- None
- Cost
- Free · Free and MIT licensed; the author takes sponsorships. Running big models on your own hardware or a rented server is the real cost.
- License
- MIT (Permissive)
- Can replace
- ChatGPT Plus, ElevenLabs, OpenRouter, Veo 3
- Last change on GitHub
- Nov 10, 2026
LocalAI risk signals (5)
- Attentionunsigned mac installer
README: "**Note:** The DMG is not signed by Apple. After installing, run: `sudo xattr -d com.apple.quarantine /Applications/LocalAI.app`", which means turning off a macOS protection for this app with an administrator command.
AI review
- Infodownloads backends at runtime
"Backend Gallery — Install/remove backends on the fly via OCI images" and "Automatic Backend Detection: LocalAI automatically detects your GPU capabilities and downloads the appropriate backend": most of the 60+ engines and all models are fetched from the internet after you start it.
AI review
- Infobuilt in agent runs commands
README on the terminal agent: "It answers questions, reads your files and runs commands on your machine, asking you to approve anything that changes state."
AI review
- Infoexample configs ship default passwords
`docker-compose.distributed.yaml` sets `POSTGRES_PASSWORD: localai`, `LOCALAI_REGISTRATION_TOKEN: "changeme"` and fixed storage keys; the file itself says "All credentials below are examples: change them for anything beyond a local test."
AI review
- Infoface and voice biometrics included
The backend table lists `face-detect.cpp` for "face detection, recognition, demographics and anti-spoofing" and `voice-detect.cpp` for speaker recognition with age, gender and emotion, so the server can identify people from photos and recordings.
AI review
Evidence is shown in its original language.
What Radar could not check about LocalAI
- How much disk space and memory it needs once the backends and models are downloaded, which the release files do not include
- Whether a given model really runs on your graphics card or only on the processor, and how fast
- Whether the macOS and Linux launchers work as smoothly as the Docker route
- Whether the many third-party backends are kept up to date and what each one downloads
- Whether the "your data never leaves your infrastructure" claim holds for every add-on and connector