Google presents Gemma 3: Open AI models with up to 27 billion parameters [German]

Author: IT BOLTWISE | Source: IT BOLTWISE | Read the full article in German

Google has recently unveiled its latest collection of open AI models called Gemma 3, which includes models with up to 27 billion parameters. This initiative aims to create a transparent and collaborative environment for artificial intelligence, allowing developers, researchers, and businesses to explore the capabilities of generative AI freely. Unlike Google's proprietary Gemini models, the Gemma models are openly accessible, providing users with the flexibility to adapt them to their specific needs.

The Gemma 3 collection features models optimized for various computing requirements, making them suitable for both startups and large enterprises. This flexibility allows organizations to run the models on their own infrastructure, reducing reliance on cloud services. Google has ensured compatibility with popular frameworks like TensorFlow, JAX, and PyTorch, and has provided extensive documentation to assist users in implementing these models effectively.

A key focus of the Gemma models is on safety, responsibility, and ethics. Google has implemented strict measures to minimize biases and prevent misuse, addressing concerns raised about open models in the past. The release of Gemma 3 also responds to the increasing demand for robust open-source models, reinforcing Google's commitment to open science and its competitiveness in the AI landscape.

[Read More (translated)]