Gemma 3 is a collection of lightweight, state-of-the-art open AI models developed by Google DeepMind, designed to deliver advanced performance on a single GPU or TPU. Built upon the research and technology that powers the Gemini 2.0 models, Gemma 3 offers multimodal capabilities, supporting both text and image inputs, and provides extensive multilingual support for over 140 languages. With model sizes ranging from 1 billion to 27 billion parameters, Gemma 3 is tailored for efficient deployment across various devices, from smartphones to workstations.