This AI Model Runs On Your Phone (With No Internet)!

· Source: Matt Wolfe · Field: Technology & Digital — Artificial Intelligence & Machine Learning, Software Development & Engineering, Emerging Technologies & Innovation · Depth: Intermediate, long

Summary

The Locally AI app enables users to run powerful open-weight AI models, specifically the new Quinn 3.5 series, directly on their iPhones without an internet connection. Released on March 2nd, the Quinn 3.5 model comes in 800 million, 2 billion, 4 billion, and 9 billion parameter variations, with the 4 billion parameter version recommended for iPhone 15 Pro or newer, and the 2 billion parameter version for iPhone 15. The app, developed by Adrian Gronden, boasts a 4.8-star rating and allows for model selection, custom instructions, temperature control, and Siri shortcuts. It supports basic text queries, brainstorming, and visual input, demonstrating offline functionality even in airplane mode. While not matching the largest cloud-based models like GPT-4, these local models offer performance comparable to or better than state-of-the-art models from a year and a half ago, ensuring user data privacy by keeping all interactions on-device.

Key takeaway

For mobile developers or privacy-conscious users considering local AI capabilities, the Locally AI app with Quinn 3.5 models offers a robust solution. You can achieve significant AI functionality, including text generation and visual analysis, entirely offline, safeguarding your data from cloud services. This capability is particularly valuable for scenarios requiring immediate, private AI assistance without internet access, such as travel or sensitive personal use cases.

Key insights

Run powerful, privacy-preserving AI models directly on your phone without an internet connection.

Principles

Method

Download the Locally AI app, select a Quinn 3.5 model variant (e.g., 2B or 4B parameters) compatible with your iPhone (iPhone 14 or newer), and utilize its text, visual, or voice modes for offline AI interactions.

In practice

Topics

Best for: Machine Learning Engineer, AI Student, Software Engineer

Related on AIssential

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by Matt Wolfe.