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

Google has quietly launched the Google AI Edge Gallery, an Android app (with an upcoming iOS version) that allows users to download and run openly available AI models from Hugging Face directly on their smartphones. This innovative app enables users to generate images, answer questions, write code, and more, all offline, utilizing the processing power of their devices without requiring an internet connection.

This approach offers an alternative to cloud-based AI models, addressing concerns about data privacy and the need for constant connectivity. The Google AI Edge Gallery provides a user-friendly interface with features like "Ask Image" and "AI Chat," which lead users to compatible AI models, including Google's Gemma 3n.

The app also includes a "Prompt Lab" where users can experiment with "single-turn" tasks such as text summarization and rewriting, offering customizable settings to fine-tune the models' performance. The app's open-source nature, under an Apache 2.0 license, allows for unrestricted use, including commercial applications.

Google acknowledges that performance may vary depending on the device's hardware and model size. More powerful devices will predictably run models faster, but the size of the model also plays a role, with larger models requiring more processing time. Google is actively seeking feedback from the developer community to improve the Google AI Edge Gallery experience, emphasizing its experimental Alpha release status.

This release signifies a growing trend toward on-device AI, offering users greater control over their data and access to AI capabilities even without an internet connection. By leveraging the power of smartphones, Google AI Edge Gallery opens up new possibilities for AI applications, making them more accessible and versatile for a broader audience.