Gemma

Gemma is Google DeepMind's family of lightweight, state-of-the-art open models, built on the same core technology as the Gemini series.

Gemma is Google DeepMind's open-weight model family (LLMs and VLMs), leveraging the same proprietary research and technology as the high-performance Gemini series. The models are released in various parameter sizes, from 270M up to 27B, and deliver state-of-the-art performance for their scale (e.g., Gemma 3 27B). Developers use Gemma to build and customize generative AI applications, supporting diverse use cases like on-device deployment (Gemma 3n) and specialized tasks (CodeGemma, PaliGemma 2). This release promotes responsible AI innovation and community-driven development.

https://ai.google.dev/gemma