Fullmoon
A privacy-focused app for running large language models directly on Apple devices. Works fully offline with optimized on-device performance.
Fullmoon is an open-source alternative to:
Fullmoon is a lightweight app designed to bring the power of large language models to your Apple devices. Whether you're using an iPhone, iPad, Mac, or Vision Pro, Fullmoon provides a seamless way to interact with LLMs without compromising your privacy.
Key Features
- Private & Local: Runs models entirely on-device with no cloud dependency, ensuring your data never leaves your device.
- Offline Capability: Works fully offline when internet is unavailable, perfect for travel or areas with poor connectivity.
- Apple Silicon Optimized: Leverages Metal 3 and Swift MLX for maximum performance on Apple hardware.
- Model Variety: Supports multiple quantized models including Llama 3.2 (1B/3B) and DeepSeek R1 variants.
- Customization: Personalize themes, fonts, and system prompts to tailor the experience to your preferences.
- Shortcuts Integration: Use local models with Apple Shortcuts for automation and workflow integration.
- TestFlight Access: Get early access to new features and models through TestFlight.
Tech / Self-Hosting
Fullmoon is currently available as an iOS/iPadOS/macOS app with no self-hosting option. The app is open source (GitHub repository) and runs exclusively on Apple devices with Apple Silicon processors. Models are quantized for efficient on-device inference.