Last week, Google launched the ML Kit GenAI application programming interface (API), enabling Android developers to create applications that utilize Gemini Nano’s capabilities. According to a document posted on its developer forum, the tech giant has also made the image description feature of its artificial intelligence (AI) model available to developers for the first time. Previously, access to the model was limited to experimental use, preventing developers from publishing apps that utilized the large language model (LLM).
Developers Can Now Use Gemini Nano to Build AI-Powered Android Apps
As first reported by Android Authority, Google has updated Gemini Nano’s Android developer page to include a support document for the new ML Kit GenAI API. This API allows developers to “harness the power of Gemini Nano to deliver out-of-the-box performance for common tasks through a simplified, high-level interface.”
The document further notes that the API is built on AICore, an Android system service, which facilitates on-device execution of Gemini Nano-like models, regardless of developers’ familiarity with model mechanics. Applications created with this AI model will function locally, utilizing the device’s system-on-a-chip (SoC).
The ML Kit GenAI API provides access to various new features, including text summarization, message proofreading, message rewriting, and the ability to add short descriptions to images. Notably, Google has announced a session at I/O 2025 titled “Gemini Nano on Android: Building with on-device gen AI,” where the company is expected to detail the model’s capabilities and guide developers on how to integrate these features into their applications.
Gemini Nano was initially introduced to developers in October 2024 as a part of the AI Edge software developer kit (SDK). However, that version was limited to experimental access, preventing app publication.
Furthermore, the original SDK supported application development exclusively for the Google Pixel 9 series, while the new API extends compatibility to all appropriate Android devices. The previous SDK was restricted to text-based features, lacking the newly available image description functionality.
