Llama macOS is a native macOS menu bar application for downloading, running, and managing large language models locally. Built around llama.cpp, it provides a straightforward way to use AI models without cloud processing. It automatically detects existing llama.cpp installations or installs a compatible binary when necessary. Users can download GGUF models from Hugging Face and receive recommendations based on their Mac's hardware. An integrated web interface supports direct conversations with installed models. The application starts a local OpenAI-compatible API server for connecting coding assistants, editors, and other tools. Models load on demand and unload when idle to conserve memory. Model configurations are optimized automatically, with optional overrides for advanced users.
Features
- Private local large language model inference powered by llama.cpp on macOS
- One-click GGUF model downloading from Hugging Face with hardware-based recommendations
- Automatic llama.cpp installation and detection of previously downloaded models
- Integrated web chat interface for interacting with locally installed AI models
- OpenAI-compatible local API server for connecting editors and external AI applications
- Automatic model loading and unloading with customizable inference configuration settings