Google unveils Gemma family of lightweight generative AI models
• Gemma is a new family of open, lightweight generative AI models from Google, built using similar technology as Gemini models.
• Two model sizes are released - Gemma 2B and Gemma 7B, with pre-trained and instruction-tuned variants of each.
• Tools and resources provided include a Responsible Generative AI Toolkit, integrations with popular frameworks, and optimized performance on NVIDIA GPUs and Google Cloud TPUs.
• Gemma models achieve state-of-the-art capabilities compared to other models of similar size.
• Free access and credits are available to help researchers and developers get started with Gemma.