Llama

About

Llama is a macOS menu bar app for running local LLMs. When you start it, it runs a local server at http://localhost:9931/v1. If you already have llama.cpp installed, Llama uses it; otherwise, it installs a prebuilt binary for your Mac. Models you've installed through llama.cpp appear in the app automatically, and you can install any GGUF model from Hugging Face. Llama also recommends models that fit your Mac's hardware. Models run entirely on your Mac, load when requested, and unload when idle. You can chat with models in the built-in WebUI, connect other apps such as coding agents, chat UIs, and editors, or use the API directly.

The app is a 4 MB native macOS app and is designed to need no configuration: models are set up with settings suited to your Mac. Models are stored in the Hugging Face cache, so they can be shared with llama.cpp and other tools. Llama is built on llama.cpp and is developed alongside it.

By default, the server is reachable only from your Mac. The "Allow network access" setting can expose it either through Tailscale, which binds the server to your Tailscale address and authenticates and encrypts connections, or to all network interfaces, which the site warns should be used only on networks you trust because the server has no password.