Google DeepMind released Gemma 4 12B, an open-source multimodal model that processed text, images, and audio natively, ran on laptops with 16 GB of RAM, nearly matched the 26B model in benchmarks, and was distributed under an Apache 2.0 license permitting commercial use.