Show HN: Gemma 4 26B A4B running on an iPhone 17 Pro via model paging

Category: ai-ml

Tags: on-device-ai, llm, apple, privacy, model-paging

Score: 6.7/10 (Innovation: 7, Technical: 7, Documentation: 3, Utility: 6)

Noema is a private local AI app for Apple devices that runs large language models like Gemma 4 26B on an iPhone using a model paging technique. It focuses on privacy by keeping models, files, and conversations on-device, with support for file search and tools. The project is interesting for demonstrating large model inference on mobile hardware without cloud dependency.

Target audience: iOS developers, AI enthusiasts, privacy-conscious users

Repository: https://noemaai.com/

View on Hacker News