Google releases Gemma 4 12B model for 16GB enterprise laptops
Google just released Gemma 4 12B, a new AI model built to run smoothly on enterprise laptops with only 16GB of memory.
It's an open-weights model released under the Apache 2.0 license and works offline, which means better security and less worry about internet issues or expensive cloud bills, pretty handy for businesses wanting more control.
Gemma processes audio and visuals directly
Gemma skips the usual encoders and processes audio and visuals directly, making it faster and easier on your device's memory.
With its massive context window (256,000 tokens) and step-by-step reasoning, it's great for tackling long documents or complex tasks.
Even though it's smaller than some big-name models, it holds its own in performance.
You can grab it from Hugging Face or plug it into tools like vLLM and llama.cpp to get started quickly.