Switzerland has entered the artificial intelligence race with Apertus, a new open-source Large Language Model developed by leading Swiss universities. The model aims to provide a transparent alternative to commercial AI systems like ChatGPT and Meta’s Llama.
Apertus, meaning “open” in Latin, was created by the Swiss Federal Institute of Technology Lausanne (EPFL), ETH Zurich, and the Swiss National Supercomputing Centre. Unlike many commercial AI models, Apertus offers complete transparency. Users can inspect its training process, source code, and datasets used for development.
The Swiss team acknowledges they cannot compete with the multi-billion dollar budgets of US companies. Instead, they focus on creating a safer and more accessible AI system for researchers and businesses. “We aim to provide a blueprint for how a trustworthy, sovereign and inclusive AI model can be developed,” said Martin Jaggi, professor of machine learning at EPFL.
Key features and capabilities
Apertus was trained on 15 trillion tokens across more than 1,000 languages, with 40 percent of data in non-English languages, including Swiss German and Romansh. The developers used only publicly available data and respected website opt-out requests during training.
The model is available in two sizes with 8 billion and 70 billion parameters. Researchers, companies, and hobbyists can download it to build chatbots, translators, and educational tools while maintaining control over their data.
Swiss industry groups welcome the initiative, particularly for its focus on data security and compliance with European regulations. However, some Swiss banks already use other AI models, making widespread adoption uncertain.