Google's new Gemma 4 12B brings multimodal AI to 16 GB laptops

Google has unveiled Gemma 4 12B, a new open, agentic multimodal AI that can run locally on laptops with even 16 GB of RAM. The model comes equipped with Multi-Token Prediction (MTP) drafters to reduce latency. It can natively process texts, images and audio through an encoder-free architecture. Google says it delivers reasoning performance close to its larger 26B model.