Google Releases Gemma 3n Open-Source AI Model That Can Run Locally on 2GB RAM

Google released a new open-source artificial intelligence (AI) model that can run locally on just 2GB of RAM. Dubbed Gemma 3n, it is built on the company’s new MatFormer architecture and uses Per-Layer Embeddings (PLE) to smartly manage storage space while remaining efficient. Notably, the model accepts multimodal input, but can only generate text output.

Tech