Google has quietly rolled out a new app that lets users run artificial intelligence models directly on their phones — no internet required.
Called AI Edge Gallery, the app is now available for Android, with an iPhone version on the way. It connects users with a wide selection of Generative AI tools, including ones that can generate images, answer questions, and even help write or edit code.
The big difference? These models run on the phone itself, not in the cloud.
That means you don’t need Wi-Fi or mobile data to use them. Everything happens locally, powered by the device’s own processor — a feature that could appeal to those who care about privacy or often work without a reliable internet connection.
Once inside the app, users can choose from various tasks like “AI Chat” or “Ask Image.” Each task links to a list of AI models suited for that job — for example, Google’s own Gemma 3n.
There’s also a feature called Prompt Lab that lets people experiment with text-based actions, such as rewriting or summarizing content. It comes with templates and settings to help users tweak how the model responds.
Google says performance will vary when using these models. Newer phones with faster chips will handle the models more smoothly, while larger models may take longer to complete tasks — especially on less powerful devices.
The app is still in an early testing phase, and Google is inviting developers to try it out and share feedback. It’s available under the Apache 2.0 license, so anyone can use it freely — even for commercial projects.
Read next: How AI and Authenticity Are Changing the Way People Search
Called AI Edge Gallery, the app is now available for Android, with an iPhone version on the way. It connects users with a wide selection of Generative AI tools, including ones that can generate images, answer questions, and even help write or edit code.
The big difference? These models run on the phone itself, not in the cloud.
That means you don’t need Wi-Fi or mobile data to use them. Everything happens locally, powered by the device’s own processor — a feature that could appeal to those who care about privacy or often work without a reliable internet connection.
Once inside the app, users can choose from various tasks like “AI Chat” or “Ask Image.” Each task links to a list of AI models suited for that job — for example, Google’s own Gemma 3n.
There’s also a feature called Prompt Lab that lets people experiment with text-based actions, such as rewriting or summarizing content. It comes with templates and settings to help users tweak how the model responds.
Google says performance will vary when using these models. Newer phones with faster chips will handle the models more smoothly, while larger models may take longer to complete tasks — especially on less powerful devices.
The app is still in an early testing phase, and Google is inviting developers to try it out and share feedback. It’s available under the Apache 2.0 license, so anyone can use it freely — even for commercial projects.
Read next: How AI and Authenticity Are Changing the Way People Search
