ProductJun 4, 2026, 02:49 AM
Google's new open source Gemma 4 12B analyzes audio, video — and runs entirely locally on a typical 16GB enterprise laptop
Summary
Google released Gemma 4 12B, an 11.95-billion-parameter open-weights model under Apache 2.0 license, optimized to run locally on enterprise laptops with 16GB VRAM. It features an encoder-free 'Unified' architecture for direct processing of raw audio and visual data, supports 256K token context window, native tool use, and step-by-step reasoning, enabling offline AI operations.
Why it matters
The model demonstrates Google's progress in small, local AI models, significant for edge computing and enterprise security scenarios.
Source links
Market reaction
Alphabet / Google
NASDAQ · GOOGL
Micron Technology
NASDAQ · MU
