Gemini Nano

Gemini Nano is Google's most efficient on-device AI model, engineered for high-end mobile chipsets (Tensor G4, Snapdragon 8 Gen 3) to deliver low-latency, private, and offline generative AI capabilities.

This is Google's specialized, compact large language model, built to run directly on a device like the Pixel 9 series. Nano executes core AI functions—summarization, smart reply suggestions, and image description via TalkBack—locally, ensuring maximum privacy and minimal latency. The model leverages dedicated mobile hardware, specifically the Neural Processing Units (NPUs) in chips like the Tensor G4, to process requests offline without sending user data to the cloud. This on-device architecture is the key: it provides near-immediate results for text-based and multimodal tasks, making powerful generative AI a standard, secure feature of the mobile experience.

https://deepmind.google/technologies/gemini/nano/