The latest Gemini Nano, a powerful on-device multi-modal model, is now available through ML Kit GenAI APIs on Pixel 10 devices. Developers can integrate features like summarization, proofreading, rewriting, and image description into their applications. The GenAI APIs aim to provide consistent quality across different model versions and device upgrades. This consistency is achieved through a rigorous evaluation pipeline using LLM-based raters, statistical metrics, and human feedback. Additionally, feature-specific LoRA adapters are trained and deployed on top of the Gemini Nano base model to meet quality standards. The newest Gemini Nano significantly improves performance, particularly in prefix speed, which is how quickly the model processes input. Benchmarks on Pixel 10 Pro show substantial increases in tokens per second for both text-to-text and image-to-text tasks. The team is dedicated to maintaining this quality assurance process as Gemini Nano continues to evolve. This approach aims to simplify the integration of Gemini Nano into Android apps, allowing developers to leverage new capabilities easily. Developers can find resources and guidance in the official documentation and sample code for implementing GenAI APIs.