Switzerland has officially entered the AI landscape with the release of Apertus, a new open-source national Large Language Model (LLM). Developed as a public infrastructure project by the Swiss Federal Technology Institute of Lausanne (EPFL), ETH Zurich, and the Swiss National Supercomputing Centre (CSCS), Apertus is designed to be a transparent and privacy-conscious alternative to proprietary models from large tech companies.
Apertus, which is Latin for "open," was trained on 15 trillion tokens across more than 1,000 languages, including Swiss German and Romansh. The model is available in two sizes: an 8-billion-parameter version for smaller-scale applications and a 70-billion-parameter version for research and commercial use.
The developers have emphasized the model's transparency, releasing comprehensive documentation, the source code of its training process, and the datasets used. This open approach is intended to foster trust and allow for greater scrutiny of the model's development and a community-driven approach to AI. The training data complies with Swiss data privacy laws, a key consideration for a model positioned as a sovereign, European-focused AI.
The release of Apertus is a significant step in the democratization of AI and a notable development in the open-source LLM ecosystem. It provides a powerful new tool for researchers, companies, and hobbyists, and it underscores the growing importance of open, transparent, and ethical AI development.