Gemma is a family of lightweight, high-performance open AI models built on the same scientific research and technological infrastructure that underpins the Gemini models. It was developed by Google DeepMind with contributions from specialized teams across Google. Its name is inspired by Gemini and the Latin word gemma, meaning precious stone, reflecting the idea of a model that is both lightweight and valuable.
The Gemma models are offered as open models, enabling broad use by researchers, developers, and organizations. Alongside the model weights, an integrated ecosystem of tools is provided, which:
- Facilitates the fast development of innovative AI applications.
- Enhances collaboration among developers and researchers.
- Promotes responsible and safe use of the models, with a focus on transparency and risk mitigation.
- Supports flexible customization for use on both personal computers and cloud environments.
Thanks to their low computational requirements and strong performance, Gemma models are particularly suitable for developing AI applications on smaller or local systems without the need for specialized hardware.
Last updated on: 15/12/2025 18:00
