Google Debuts App for Local AI Processing

Google has quietly introduced a new Android app, "Google AI Edge Gallery," enabling users to run artificial intelligence models directly on their smartphones without an internet connection. An iOS version is expected soon.

This experimental Alpha release allows users to browse, download, and execute a range of AI models from Hugging Face, a popular platform for open-access AI. The app facilitates diverse tasks, including image generation, question answering, and code writing/editing, with all processing handled locally on the device's hardware, eliminating reliance on cloud servers.

Running AI models locally offers significant benefits, particularly for privacy-conscious users and those requiring offline access. By keeping data on the phone, users avoid sending sensitive information over the internet.

"Google AI Edge Gallery" is available for download on GitHub with specific installation instructions. Its home screen features shortcuts for various AI tasks like "Ask Image" and "AI Chat," letting users explore compatible models such as Google's Gemma 3n. The app also includes a "Prompt Lab" for single-turn tasks like summarizing or rewriting text, with customizable templates for fine-tuning model behavior.

Google notes that performance will vary based on hardware and model size; high-end smartphones will process models faster, while larger models may require more time.

Licensed under Apache 2.0, "Google AI Edge Gallery" is open for both commercial and non-commercial use. Google is actively seeking feedback from the developer community to refine the app as it evolves.