Particle.news
Download on the App Store
3 ARTICLES
·
yesterday

Google DeepMind Previews Gemma 3n for Mobile-First, Multimodal AI

The new AI model supports real-time, on-device processing across phones, tablets, and laptops, with advanced efficiency and multimodal capabilities.

Image
Image
Image
Image

Overview

  • Gemma 3n is optimized for on-device AI, enabling real-time performance on mobile devices while reducing reliance on cloud processing.
  • The model introduces innovations like Per-Layer Embeddings, allowing 5B and 8B parameter models to run with minimal memory usage comparable to smaller models.
  • Gemma 3n supports multimodal inputs, including audio, text, images, and video, with enhanced multilingual capabilities for languages such as Japanese, German, and Spanish.
  • Developers can access Gemma 3n through Google AI Studio for cloud-based use or Google AI Edge for on-device integration starting today.
  • Google collaborated with Qualcomm, MediaTek, and Samsung to optimize Gemma 3n's mobile-first architecture, while also announcing MedGemma for healthcare AI and SignGemma for sign language translation.