Google Gemma

by

Google Unveils Gemma 3n and Specialized AI Models at I/O 2025

Gemma 3n: On-Device AI Power
At Google I/O 2025, Google introduced Gemma 3n, a versatile AI model optimized for local device operation. The preview version launched Tuesday can process:

  • Text

  • Audio

  • Images

  • Videos

 

Key Advantages of On-Device AI:

  • Lower operational costs than cloud-based models

  • Enhanced privacy by keeping data local

  • Ability to run on devices with under 2GB RAM

“Gemma 3n shares Gemini Nano’s architecture but delivers even better performance,” said Gemma Product Manager Gus Martins during his keynote.

Specialized AI Offerings

MedGemma for Healthcare
Google’s Health AI program launched MedGemma, its most advanced open model for medical applications. The multimodal system excels at analyzing:

  • Medical texts

  • Clinical images

“Developers can adapt MedGemma for various health apps,” Martins explained.

 

SignGemma for Accessibility
Coming soon, SignGemma will translate sign language to text with initial focus on:

  • American Sign Language

  • English

“This breakthrough model will empower deaf communities and developers,” Martins added.

Developer Reception and Concerns

While developers have downloaded Gemma models tens of millions of times, some criticize its non-standard licensing terms for creating commercial use uncertainties

Leave a Comment