Fullmoon - A billion parameters in your pocket - chat with private and local large language models

Fullmoon A billion parameters in your pocket - chat with private and local large language models

What is Fullmoon?

Fullmoon represents a breakthrough in private AI computing, offering users the ability to run large language models directly on their Apple devices. The application is specifically optimized for Apple silicon and leverages Metal 3 graphics technology to ensure efficient model execution.

Built on the Swift MLX array framework, Fullmoon supports multiple platforms including iOS, iPadOS, macOS, and visionOS. The app comes with two pre-configured models: Llama-3.2-1B-Instruct-4bit (0.7 GB) and Llama-3.2-3B-Instruct-4bit (1.8 GB), both designed to operate efficiently with 4-bit precision.

Features

Offline Operation: Works completely without internet connection
Local Processing: Runs models directly on device for privacy
Apple Silicon Optimization: Specially optimized for Apple's custom chips
Cross-Platform Support: Works on iOS, iPadOS, macOS, and visionOS
Customization Options: Adjustable themes, fonts, and system prompts
Shortcut Integration: Compatible with Apple's Shortcuts app

Use Cases

Private AI conversations without internet connectivity
Secure local language processing
Integration with Apple device automation
Offline language model deployment
Personal AI assistant with privacy focus

FAQs

What are the minimum system requirements for Fullmoon?

Fullmoon requires an Apple device with Apple silicon chip and Metal 3 graphics support.
How much storage space do the models require?

The Llama-3.2-1B-Instruct-4bit model requires 0.7 GB, while the Llama-3.2-3B-Instruct-4bit model needs 1.8 GB of storage space.
Can I use Fullmoon without an internet connection?

Yes, Fullmoon works completely offline as it runs the language models locally on your device.