Google has quietly launched a groundbreaking app called AI Edge Gallery, allowing users to run open AI models directly on their smartphones—without an internet connection. Currently available for Android (with iOS coming soon), the app taps into models from Hugging Face, enabling tasks like image generation, code editing, and answering questions entirely on-device.
Privacy and offline access are the key selling points. Unlike cloud-based AI services (like Gemini or ChatGPT), which require sending data to external servers, AI Edge Gallery keeps everything local. This makes it ideal for sensitive tasks or areas with spotty connectivity. The app supports models like Google’s Gemma 3B, and users can even integrate custom models via LiteRT.
The interface is simple: a home screen with shortcuts for AI Chat, Ask Image (analyze photos), and Prompt Lab (for text tasks like summarization). Performance depends on your phone’s hardware—newer devices handle larger models faster. Google warns it’s still an experimental alpha, available only via GitHub for now, but it’s open-source (Apache 2.0 license) and free for commercial use.
Analysts note this aligns with Google’s broader AI shift. David Hunter of Local Falcon highlights how AI Overviews are changing search: businesses now need quality content over keyword stuffing to rank. Meanwhile, Google keeps integrating AI elsewhere, like Gmail’s auto-summaries and ads in AI Search.
For developers, the app is a sandbox to test offline AI’s limits. For everyday users? A glimpse at a future where your phone’s brain works solo—no Wi-Fi required. Just don’t expect ChatGPT-level power yet. As one reviewer put it: “Local AI is wild… until your phone starts sweating.”
Also read: Google’s new update you should know about for your privacy
