Google Launches Gemma, a New Open-Source AI Model Family
Google introduces Gemma, a lightweight, state-of-the-art family of open AI models derived from Gemini technology. The release includes robust developer toolkits and emphasizes responsible AI development for researchers and commercial use.
Google introduces Gemma, a new family of lightweight, state-of-the-art open models built from the same research and technology used to create the Gemini models. Developed by Google DeepMind and other teams across the company, Gemma extends Google's long history of contributing open-source innovations like TensorFlow, BERT, and JAX to the broader AI community.
The tech giant releases model weights in two sizes, Gemma 2B and Gemma 7B, with both pre-trained and instruction-tuned variants available today. To support developer innovation, Google provides comprehensive toolchains for inference and fine-tuning across major frameworks like JAX, PyTorch, and TensorFlow through native Keras 3.0. The models integrate seamlessly with popular tools such as Hugging Face, Colab, and Kaggle notebooks.
Gma emphasizes responsible AI development by accompanying the model weights with a new Responsible Generative AI Toolkit designed to guide the creation of safer AI applications. The models run efficiently on laptops, workstations, or Google Cloud, offering optimized performance across hardware platforms like NVIDIA GPUs and Google Cloud TPUs. Furthermore, permissive terms of use allow organizations of all sizes to utilize Gemma for responsible commercial purposes.