Google quietly released an app that lets you download and run AI models locally

Last week, Google quietly released an app with which users can perform a series of openly available AI models of the AI DEV platform cuddle on their phones.
Called Google Ai Edge GalleryThe app is available for Android and will soon come to iOS. This allows users to find, download and implement compatible models that generate images, answer questions, write and edit code and more. The models are performed offline without needing an internet connection that use the processors of supported phones.
AI models that run in the cloud are often more powerful than their local counterparts, but they also have their disadvantages. Some users can be wary of sending personal or sensitive data to an external data center, or want models available without finding a Wi-Fi or mobile connection.

Google Ai Edge Gallery, who calls Google an ‘experimental alpha release’, can be downloaded from Github by following These instructions. The home screen shows shortcuts to AI tasks and possibilities such as “Ask Image” and “Ai Chat”. Ticking for possibilities picks up a list with models that are suitable for the task, such as Google’s Gemma 3n.
Google Ai Edge Gallery also offers a “prompt lab” users who can use to start “single -turn” tasks that are powered by models, such as summarizing and rewriting text. The prompt lab comes with different task templates and configurable settings to refine the behavior of the models.
Your mileage can vary in terms of performance, warns Google. Modern devices with more powerful hardware will perform predictable models faster, but the model size is also important. Larger models will take more time to complete a task – for example, answer a question about an image – than smaller models.
The inviting members of the Google developer community to give feedback about the Google Ai Edge Gallery experience. The app is under an Apache 2.0 license, which means that it can be used without restriction in most contexts – commercially or otherwise.