Google DeepMind Releases Gemma 3n: A Compact, High-Efficiency Multimodal AI Model for Real-Time On-Device Use

Google announces in preview Gemma 3n, the latest version of its open AI model capable of operating on a single GPU, Gemma 3n would be a powerful, efficient and mobile-first IAGoogle unveiled Gemma 3n, a new addition to its family of "open" AI models designed to work effectively on smartphones, laptops and tablets with only one GPU. Gemma 3n would be able to handle text, images and, ultimately, audio and video...

Google publishes Gemma 3n, a multimodal SLM adapted for on-device inference. The model takes advantage of a brand new architecture developed by DeepMind.

Google DeepMind Releases Gemma 3n: A Compact, High-Efficiency Multimodal AI Model for Real-Time On-Device Use

Researchers are reimagining how models operate as demand skyrockets for faster, smarter, and more private AI on phones, tablets, and laptops. The next generation of AI isn’t just lighter and faster; it’s local. By embedding intelligence directly into devices, developers are unlocking near-instant responsiveness, slashing memory demands, and putting privacy back into users’ hands. With mobile hardware rapidly advancing, the race is on to build co…