Google has launched a new app called ‘AI Edge Gallery’ that allows Android phones to run powerful AI models directly on the device, without needing an internet connection. This means users can use AI features like creating images, answering questions, and writing code completely offline (without connecting to the cloud).
This new app integrates models from Hugging Face, a prominent AI repository, facilitating on-device execution for enhanced privacy, faster processing, and offline functionality.
One of the key models used is Gemma 3 1B, a compact yet powerful language model designed for mobile and web platforms. At just 529MB, Gemma 3 1B can process up to 2,585 tokens per second, enabling rapid content generation and interaction. Notably, this model supports various applications, including personalized content creation, intelligent responses in messaging apps, and document analysis.
The app also leverages Google’s AI Edge platform, containing tools like MediaPipe and TensorFlow Lite, to optimize model performance on mobile devices. These tools facilitate the deployment of AI models by providing efficient inference capabilities and hardware acceleration, ensuring smooth operation even on devices with limited resources.
The interface of the app also includes categories like ‘AI Chat’ and ‘Ask Image’, which guide users to relevant tools. There is also a ‘Prompt Lab’ included, which acts as a sandbox where users can test and refine single-turn prompts.
Additionally, by enabling on-device AI processing, the Mountain View-headquartered company addresses concerns related to data privacy and latency. Users can perform complex AI tasks without transmitting data to external servers, reducing potential security risks and ensuring faster response times.
The app is currently in an ‘experimental Alpha release’, which means it’s still being developed, and the company is asking developers and users for feedback. It’s open-source under the Apache 2.0 license, so anyone can use it freely, even for commercial purposes. Apart from Android, the tech giant will also soon launch the app on iOS.
However, the experience may vary depending on a phone’s hardware capabilities. While newer devices with high-performance chips can run models quickly and smoothly, older phones may struggle with lag, especially when using larger models. Google notes that model size impacts processing speed and user experience, so selecting a lighter model may be the best option for users on mid-range or older devices.
Despite these developments, Google’s AI journey is no stranger to controversies and challenges. For example, just last week, the US Department of Justice launched a civil antitrust investigation into Google’s licensing agreement with AI startup Character.AI. The inquiry originates from concerns that the deal may have been intentionally structured to sidestep federal merger review requirements.
