Google to give app devs access to Gemini Nano for on-device AI

The rapid expansion of Generative AI has changed the style of design products of Google and other tech giants, but most of the AI ​​-used features you use are operating on remote servers with a ton processing power. Your phone There is very little strengthBut Google is ready to give developers some important new mobile AI tools. Next week in I/O, Google will likely announce a new set of APIS to take advantage of the developers’ capabilities for the Gemino Nano’s capabilities for the AI.

Google quietly published documents for developers on major new AI features. According to Android AuthorityThe update of the ML Kit SD will include the support of the API for the On -Device Generative AI features through Gemini Nano. This experimental edge is made on AI Core like AI SDK, but it is included in a current model that is a combination of default features that should be easy to implement the developers.

Google says ML Kit’s Geni APIS apps will enable summary, proof reading, rewriting, and image details without sending data to the cloud. However, Gemini Nano does not have the same strength as a cloud -based version, so expect some limits. For example, Google notes that the summary can have only three built -in points, and the image explanation will only be available in English. The quality of the output can also vary based on the version of Gemini Nano on the phone. Standard version (Gemini Nano XS) is about 100MB size, but Gemini Nano XXS As seen on Pixel 9A A quarter of the size. It is just the text and has a small window of context.

Not all versions of Gemini Nano are made equal.

Credit: Ryan Whiteom

This move is usually good for Android because the ML Kit works on Google’s pixel line on the outside. Although Pixel devices are widely used by Gemini Nano, several other phones have already been developed to run this model, including OnePlus 13, Samsung Galaxy S25, and Xiaomi 15. Since more phones increase the support of Google’s AI model, the developers will be able to target these devices with the properties of generating AI.