Responsible AI Models
Gemma is designed with a focus on responsible AI principles, using automated techniques to filter out personal and sensitive data from training sets. Through extensive fine-tuning and reinforcement learning from human feedback, Gemma models are aligned with responsible behaviors. Robust evaluations, including manual and automated testing, are outlined in the Model Card to understand and reduce risk profiles.
In addition, a new Responsible Generative AI Toolkit is being released alongside Gemma to help developers prioritize building safe and responsible AI applications. The toolkit includes safety classification, debugging tools, and guidance based on best practices from Google’s experience.
Optimized Technology
Gemma models can be fine-tuned and run across a variety of frameworks, devices, and hardware platforms. It supports multi-framework tools, cross-device compatibility, and is optimized for Google Cloud, as well as partnering with NVIDIA for industry-leading performance.
Free Credits for Research and Development
Gemma is available for free access in Kaggle, a free tier for Colab notebooks, and offers $300 in credits for first-time Google Cloud users. Researchers can also apply for Google Cloud credits of up to $500,000 to accelerate their projects.
To get started with Gemma and access quickstart guides, visit ai.google.dev/gemma.
As new variants for diverse applications are introduced, there will be opportunities to connect, learn and build with Gemma in the coming weeks, offering exciting possibilities for creators.