Repolode
Submit
ggml-org

ggml-org/Llama-macOS

A macOS menu bar app for running local LLMs, built on llama.cpp, with a local API server and built-in WebUI.

Open-source alternative to:LM StudioOllama

View on GitHubWebsite
Stars
1,652
Forks
130
Open issues
10
Last commit
2 days ago

What is it?

Llama is a macOS menu bar app that runs large language models entirely on your Mac. When launched, it starts a local server at localhost:9931 with an OpenAI-compatible /v1 endpoint, so other apps, coding agents, and editors can connect to it. If llama.cpp is already installed, the app uses it; otherwise it installs a prebuilt binary for your Mac.

The app can download any GGUF model from Hugging Face and recommends models that fit your Mac's hardware. Models share the Hugging Face cache with llama.cpp and other tools, load on request, and unload when idle to free memory. Per-model settings are generated automatically and can be overridden through a user INI file.

By default the server is reachable only from the local machine. Settings let you expose it over Tailscale or across the local network, though the server has no password in that mode.

Key features

  • 01Runs models entirely on the Mac, with no data leaving the device.
  • 02Provides a local OpenAI-compatible API at localhost:9931/v1.
  • 03Includes a built-in WebUI for chatting with models.
  • 04Auto-configures model settings and recommends models for your hardware.
  • 05Loads models on demand and unloads them when idle.
  • 06Shares the Hugging Face model cache with llama.cpp.

Use cases

  • Running private chat or coding assistants without sending prompts to a cloud service.
  • Pointing local coding agents or editors at a self-hosted model on a Mac.
  • Testing GGUF models from Hugging Face on Apple hardware without manual setup.
  • Accessing a model running on a Mac from other devices over Tailscale.

Quick start

  1. 1

    Install via Homebrew

    brew install --cask llama-app
  2. 2

    Or download the app from the Releases page

  3. 3

    Start Llama; it runs a local server at http://localhost:9931/v1

Things to know

  • macOS only.
  • The server has no password by default when network access is enabled, so it should only be exposed on trusted networks.

Topics

#llm#local-llm#llama-cpp#macos#gguf#openai-compatible-api#menu-bar-app#ai#llms#swift

FAQ

What is Llama-macOS?
A macOS menu bar app for running local LLMs, built on llama.cpp, with a local API server and built-in WebUI. Llama is a macOS menu bar app that runs large language models entirely on your Mac. When launched, it starts a local server at localhost:9931 with an OpenAI-compatible /v1 endpoint, so other apps, coding agents, and editors can connect to it. If llama.cpp is already installed, the app uses it; otherwise it installs a prebuilt binary for your Mac.
Is Llama-macOS free to use? What license does it have?
Llama-macOS is open source under the MIT license. Check the license terms before using it in commercial products.
Is Llama-macOS still maintained?
Actively maintained. The last commit to Llama-macOS was on Oct 18, 2026.
What can I use Llama-macOS for?
Running private chat or coding assistants without sending prompts to a cloud service. Pointing local coding agents or editors at a self-hosted model on a Mac. Testing GGUF models from Hugging Face on Apple hardware without manual setup. Accessing a model running on a Mac from other devices over Tailscale.
How do I install Llama-macOS?
Install via Homebrew: brew install --cask llama-app Or download the app from the Releases page Start Llama; it runs a local server at http://localhost:9931/v1
Where can I get Llama-macOS?
The source code and installation instructions are on GitHub: https://github.com/ggml-org/Llama-macOS

Summary generated by AI from the project's README. GitHub stats refresh daily.

Related projects