Introducing Gemma 3

In this brand new technical blog post, we dive into GeMMa 3, a collection of lightweight, state-of-the-art open models designed to run on devices, specifically phone and laptop hardware. These models come with advanced text and visual reasoning capabilities and are supported by quantized versions for faster performance and reduced computational requirements. Gemma 3 includes ShieldGeMMa 2, a powerful image safety checker built on the Gemma 3 foundation that provides ready-made solutions for responsible AI development. Gemma 3’s open architecture allows developers to customize its safety labels for their own safety needs and to leverage existing workflows, with SEA-LION v3 as an example of a community-created model to foster communication across Southeast Asia, BgGPT demonstrating the power of Gemma to support diverse langauges, OmniAudio showcasing the potential of on-device AI for everyday devices, and Google Cloud credit options available for academic researchers to accelerate their work. Overall, GeMMa 3 is designed to provide developers with an easy-to-use set of models for building, deploying, and scaling AI applications, making it a valuable resource for anyone looking to explore and build high-quality AI.