Apertus AI

Apertus is a groundbreaking open-source large language model (LLM) developed as part of the Swiss AI Initiative, a collaborative effort led by EPFL, ETH Zurich, and the Swiss National Supercomputing Centre (CSCS). It represents Switzerland’s first fully open, transparent, and multilingual large-scale language model designed to foster innovation, transparency, and sovereignty in AI. Apertus is released under a permissive open-source license with full access to model weights, training data, and documentation, enabling researchers, developers, and organizations worldwide to build upon it for various applications such as chatbots, translation, education, and more.

Key Features

• Complete openness: full public release of model architecture, weights, training data, and recipes to ensure transparency and community trust.
• Multilingual support with training on over 15 trillion tokens across more than 1,000 languages, including underrepresented languages like Swiss German and Romansh.
• Trained on public infrastructure using green energy at the Swiss National Supercomputing Centre.
• Designed for broad usability in research, education, commercial, and societal applications with regular updates and academic collaboration.

Use Cases

• Development of chatbots, automated translation services, AI educational tools, and domain-specific language models.
• Research projects requiring transparent, customizable AI models aligned with ethical and regulatory standards.
• Businesses and startups looking for a sovereign, trustworthy open-source AI foundation.