Google releases Gemma 4 12B, an AI model designed to run on laptops

In short: Google has released Gemma 4 12B, a new AI model that is small enough to run on many consumer laptops.

What happened

Google announced a new model in its Gemma 4 family called Gemma 4 12B. It sits between the smallest Gemma models made for phones and the larger versions that usually need specialized, expensive hardware.

Google says Gemma 4 12B can run “locally,” meaning on your own computer instead of on a company’s servers. The company says you can run it if your machine has 16GB of memory, either regular system memory (RAM) or graphics memory (VRAM). The download size for the model files is just under 18GB.

Gemma 4 12B also includes two technical changes meant to help it perform well without needing as much memory. One is Multi-Token Prediction, which helps the model draft likely next words faster (like writing a few possible next lines in pencil before choosing the best one). The other is a simpler way to handle images and audio, so the model needs less extra processing when you give it a photo or a sound clip.

People can try the model through tools such as LM Studio and Google AI Edge Gallery, or download the model files from Kaggle and Hugging Face.

Why it matters

Running an AI model on a laptop can be cheaper and more private, since your prompts and files do not have to leave your device. It also gives developers and small teams more options if they cannot afford high-end AI hardware.

Source: Arstechnica