Google has launched its new AI Edge Gallery, an Android app designed to let you run advanced generative AI models directly on your smartphone. While the iOS version is on the horizon, those with compatible Android devices can get hands-on experience right now.
Built with developers, researchers, and tech enthusiasts in mind, the app lets you experiment with open-source models like Gemma 3, Gemma 3n, and Qwen 2.5 without constantly relying on cloud services. Whether you’re interested in chatting, image processing, or prompt-based responses, the integrated Prompt Lab offers a suite of options. The models range from around 500 MB to 4 GB, and even support testing of custom LiteRT-format models.
Optimised for Android 10 and above with a minimum of 6GB of RAM and advanced chipsets, the app uses the LiteRT runtime to ensure efficient on-device inference. If you’ve ever struggled with slow response times or concerns over data privacy in cloud-based systems, this solution might be just what you need. For a quick start, simply visit the “Releases” section of the Google AI Edge Gallery GitHub repository, download the .apk file, and install it.
Among its standout features are image-based Q&A, text summarisation, code generation, and smooth multi-turn conversations. Real-time performance metrics such as Time to First Token (TTFT) and decoding speed add a transparent, data-driven edge to your AI explorations, all while ensuring your data stays local and secure.
Google’s move with the AI Edge Gallery is a practical demonstration of on-device generative AI and the capabilities of its LLM Inference API. Released under the Apache 2.0 license, the project invites community input and collaboration. Early feedback praises its privacy-centric approach and offline capabilities, even as some users note size restrictions and a current lack of voice interaction. Look forward to enhancements like iOS compatibility, real-time voice features, and improved hardware acceleration.
