Gemma 3An open-source AI model optimized for single-GPU performance, supporting multimodal inputs and over 140 languages.

Gemma 3 is a collection of lightweight, state-of-the-art open models designed to run on devices, particularly optimized for single-GPU performance. It supports multimodal inputs and over 140 languages. The model comes in various sizes (1B, 4B, 12B, and 27B), allowing developers to choose the best fit for their hardware and performance needs. Gemma 3 offers advanced text and visual reasoning capabilities, a 128k-token context window, and function calling for complex tasks. It also includes quantized versions for faster performance and reduced computational requirements. The model is part of Google's commitment to making useful AI technology accessible and builds upon the same research and technology that powers their Gemini 2.0 models. Gemma 3 is designed to enable developers to create AI applications that can run directly on devices such as phones, laptops, and workstations. Gemma 3 delivers state-of-the-art performance for its size, outperforming other models like Llama3-405B, DeepSeek-V3, and o3-mini in preliminary human preference evaluations. It allows for global applications with out-of-the-box support for over 35 languages and pretrained support for over 140 languages. The model enables the creation of AI-driven workflows using function calling and structured output. The development of Gemma 3 included rigorous safety protocols, such as extensive data governance, alignment with safety policies via fine-tuning, and robust benchmark evaluations. The Gemma family of open models has seen significant adoption, with over 100 million downloads and a vibrant community that has created more than 60,000 Gemma variants. Gemma 3's capabilities make it suitable for developers looking to create engaging user experiences that can fit on a single GPU or TPU host.