Generative artificial intelligence advancements are pushing mobile technology toward a major transformation. At present, AI operations mostly run on powerful remote servers, but Google plans to give developers tools to use on-device AI capabilities. Enthusiasm mounts for Google’s future I/O event, which should reveal a new set of APIs that enable Android users to take advantage of the Gemini Nano model right on their devices. The initiative will deliver advanced AI capabilities to users while improving privacy measures and speeding up operations by minimizing dependency on cloud-based processing.

New information from Google’s developer documentation reveals details about future AI capabilities. ML Kit SDK will soon receive an update that introduces API support for Gemini Nano-powered on-device generative AI features, according to Android Authority. The framework builds on AI Core, similar to the experimental Edge AI SDK, yet provides a more streamlined implementation approach by integrating existing models with well-defined features for developers. The latest update demonstrates Google’s commitment to making AI integration simpler and more accessible for developers working on mobile platforms.

Google’s documentation states that applications will be able to execute multiple essential tasks on-device through the new ML Kit GenAI APIs without requiring sensitive user data to be sent to cloud services. The features integrated into this system enable text summarization while also providing proofreading and rewriting capabilities, along with image description. The reduced processing power available on mobile devices imposes limitations on the Gemini Nano version installed directly on them. The functionality of summaries will be restricted to three bullet points, while initial image descriptions will only be available in the English language. The performance of AI outputs through Gemini Nano may fluctuate with different versions installed on phones. Gemini Nano XS occupies around 100MB, but Gemini Nano XXS, found on Pixel 9a devices, occupies only 25MB while offering text-only responses with limited context window capabilities.

Google’s strategy benefits the entire Android system because the ML Kit SDK works with multiple device brands outside of Google’s Pixel range. The Gemini Nano model is extensively used in Pixel phones, but device makers like OnePlus with the 13, Samsung with the Galaxy S25, and Xiaomi with the 15 are also developing their phones to support this AI model. With support for Google’s on-device AI model expanding across Android phones, developers will be able to reach broader audiences through generative AI features, which will drive innovation and deliver smarter mobile experiences across different brands.

Android app developers who want to incorporate on-device generative AI have had limited options available to them. Developers can access Google’s experimental AI Edge SDK, which enables Neural Processing Unit (NPU) operations, but it remains restricted to the Pixel 9 series and focuses on text processing. Companies such as Qualcomm and MediaTek offer APIs to manage AI workloads, but the wide range of device-specific features and functions introduces risks for developers planning long-term AI projects. Developing custom AI models demands extensive knowledge of generative AI systems. These new APIs will make local AI implementation both quicker and easier to access for many more developers.

This advancement marks an essential step towards more accessible AI integration in everyday life, although on-device AI models still fall short compared to cloud-based systems. A great number of users prefer to keep their personal data secure and private by processing it locally instead of transmitting it to remote servers. Google’s Pixel Screenshots alongside Motorola’s local notification summarization on the high-end Razr Ultra foldable demonstrate the advantages of local processing compared to cloud-based methods on the base Razr model. The implementation of standardized APIs around Gemini Nano will provide essential consistency for mobile AI development. The successful deployment of Gemini Nano relies on Google and other Original Equipment Manufacturers (OEMs) working together to achieve broad support across Android devices while recognizing that some companies may opt for different solutions, and older or less powerful phones could fail to meet local AI execution requirements.