Europe 2026
Gemini Nano on device — Florina Muntenescu & Oli Gaymond, Google DeepMind
Florina Muntenescu , Oli Gaymond
Overview
This talk introduces Gemini Nano, Google's on-device AI model for Android, and the ML Kit GenAI APIs that provide access to it. The presentation emphasizes the benefits of on-device processing, including enhanced privacy, offline capabilities, and cost savings, while also outlining hybrid approaches that leverage cloud models when necessary. The goal is to provide developers with a comprehensive offering for building intelligent experiences across a range of Android devices.
Who should watch
- AI Engineers
- Product Managers
- Developers building AI-powered features for Android
- Those interested in on-device AI and its applications
- Builders looking for solutions for privacy-sensitive data processing and offline functionality
Key takeaways
- On-device AI processing offers significant advantages such as enhanced privacy, offline functionality, and no inference costs.
- Gemini Nano is Google's efficient on-device model optimized for Android, accessible via ML Kit GenAI APIs.
- ML Kit GenAI APIs support tasks like summarization, proofreading, and rewriting, with the Prompt API being particularly powerful for natural language requests.
- The Prompt API currently accepts text and image input and provides text output, making it versatile for various use cases.
- Gemini Nano is available on flagship devices, but hybrid approaches using Firebase AI logic can extend reach to other devices by falling back to cloud inference.
- AI Core manages Gemini Nano at a system level, allowing multiple apps to share the model and optimizing performance and battery usage.
- While on-device models have battery implications, system-level optimizations and user-centric feature value help mitigate concerns.
- Future updates will include an embedding API to facilitate RAG-like solutions and similarity searches.
Notable quotes
*On-device processing means sensitive data like banking information doesn't have to leave the device.*
*We're doing the same thing for on device, right? So we get the models to the device, we make sure they're optimized.*
*The Gemini Nano models are only available on like Pixel 9, Pixel 10 kind of like that generation of devices.*
Unofficial community note. Prefer the recording for nuance.