Google Introduces Open-Source AI Models Gemma 2B, 7B for Smaller Tasks

Google released the open-source AI models based on the ones responsible for creating Gemini models called Gemma, starting today.

Gemma is divided into two sizes: Gemma 2B and Gemma 7B, both equipped with pre-trained and instruction-tuned variants.

Google Releases Gemma for Responsible AI Development

Google DeepMind and other Google teams develop Gemma. The open models are also built using the same technology to develop the Gemini models.

In addition, Google also released a new Responsible Generative AI Toolkit that will guide users in creating safe applications using Gemma. Both models can run on a laptop, workstation, or Google Cloud through Vertex AI and Google Kubernetes Engine.

"Gemma surpasses significantly larger models on key benchmarks while adhering to our rigorous standards for safe and responsible outputs," Google announced.

Google Designs Gemma With AI Principles

Google also unveiled that Gemma is built using the company's AI principles. The company used automated techniques to omit personal information and other sensitive data from training sets.

"To understand and reduce the risk profile for Gemma models, we conducted robust evaluations including manual red-teaming, automated adversarial testing, and assessments of model capabilities for dangerous activities," Google explained.

Moreover, users can customize Gemma models based on their specific application needs. The open models currently support multi-framework tools like JAX, Hugging Face Transformers, and more.

It can also run across devices from laptops to IoT, allowing users to explore its AI capabilities regardless of the device.

Gemma is offering free access using Kaggle, a free tier for Colab notebooks, and $300 in credits for first-time Google Cloud users. Google also assured that the research and development community is welcome to utilize the open models.