Made for reactive and low imprint local inference, Gemma 3n allows a new wave of smart and traveling applications.
Capacities
Multimodal understanding
Analysis and responds to combined images and text – with video and audio to come soon.
-
Confidentiality first, offline
Build intelligent and interactive features that first put the confidentiality of users and work reliably offline.
-
Optimized performance on devices
Mobile architecture first, with a considerably reduced memory imprint. Co-designed by Google mobile hardware teams and industry leaders.
-
Use of dynamic resources
4B imprint of active memory with capacity to create submodels for quality compromises of quality.
Start building with Gemma 3n
