Google’s AI Edge Gallery app turns your phone into a test bench for on‑device AI models

Google is quietly expanding a new tool for developers and AI enthusiasts: the AI Edge Gallery app, which lets you download and run AI models directly on your phone or computer instead of relying on cloud servers. Initially introduced as a GitHub preview, the project has since evolved into an open beta on Google Play and added iOS distribution, along with a steady stream of new models and experimental features.

Inside the app, users can browse a catalog of supported open models, pull them down to a local device, and try practical workflows such as image recognition or text generation entirely on‑device. Built‑in benchmarking makes it easy to compare models for speed and quality, and to see how well they behave in low‑connectivity or privacy‑sensitive scenarios where cloud inference is less desirable.

Recent Android updates layer on more advanced capabilities, including experimental support for the Model Context Protocol, local notifications, and persistent chat sessions — the kind of plumbing needed for full‑fledged AI agents that live on your phone. With industry interest in privacy‑friendly, low‑latency AI growing fast, AI Edge Gallery is emerging as a handy sandbox for anyone who wants to understand what “on‑device AI” can actually do in the real world.