Google Launches ‘Gemma’ Open AI Models for Developers

Gemma header width 1200 format webp

Google has unveiled Gemma, a new suite of open AI models, designed to empower developers and researchers with state-of-the-art tools for building AI applications responsibly.

These models are developed by Google DeepMind and various Google teams, drawing inspiration from the Gemini models and named after the Latin word for “precious stone,” said the company on Thursday.

Gemma models, available globally, come in two sizes: Gemma 2B and Gemma 7B, both offered in pre-trained and instruction-tuned variants. Google emphasizes the models’ lightweight design, making them accessible for use on developers’ laptops and desktops, as well as Google Cloud platforms.

“We’re releasing model weights in two sizes: Gemma 2B and Gemma 7B. Each size is released with pre-trained and instruction-tuned variants,” Google stated, highlighting the models’ adaptability and performance. The company also introduced a Responsible Generative AI Toolkit alongside Gemma, aiming to guide developers in creating safer AI applications.

Gemma models are optimized for a wide range of AI hardware platforms, including NVIDIA GPUs and Google Cloud TPUs, ensuring top-tier performance. They are compatible with major frameworks like JAX, PyTorch, and TensorFlow through native Keras 3.0, and integrate with popular tools such as Hugging Face, MaxText, NVIDIA NeMo, and TensorRT-LLM.

“Optimization across multiple AI hardware platforms ensures industry-leading performance, including NVIDIA GPUs and Google Cloud TPUs,” Google added, underscoring the models’ efficiency and scalability.

Designed with Google’s AI Principles in mind, Gemma models incorporate automated techniques to filter out sensitive data from training sets and have undergone extensive fine-tuning and evaluations to align with responsible behaviors. The Responsible Generative AI Toolkit provides resources for safety classification, model debugging, and best practices for model builders.

“Responsible by design, Gemma is designed with our AI Principles at the forefront,” Google explained, detailing the comprehensive measures taken to ensure the models’ safety and reliability.

Gemma’s launch is accompanied by free access offers on Kaggle, a free tier for Colab notebooks, and $300 in credits for new Google Cloud users, with researchers eligible for up to $500,000 US in credits to support their projects.

P.S. Help support us and independent media here: Buy us a beer, Buy us a coffee, or use our Amazon link to shop.