Google has unveiled a new series of open-source AI models named Gemma, derived from the research behind the Gemini model suite. Gemma models like 2B and 7B, known for their lightweight design and superior performance compared to similar open models, are now available for developers and students worldwide.
These models, which can be tailored to specific applications using provided data and tools like NVIDIA NeMo and HuggingFace, are compatible with popular frameworks such as PyTorch and TensorFlow. Gemma also supports flexible deployments with external sources like Nvidia systems, ensuring optimized performance.
Google is emphasizing the ethical use of AI with the release of Gemma, along with the Responsible Generative AI Toolkit. This toolkit offers safety protocols for model builders, ensuring the responsible application of AI technology in various fields.
Furthermore, Google is offering financial incentives to new users, including $300 in credits for Google Cloud and up to $500,000 in credits for researchers. Free access to Gemma models is available through platforms like Kaggle and Colab notebooks.
The release of Gemma models is set for February 21, with Google promising a series of events and initiatives in the near future to engage users in learning and building with the new AI technology.