Switzerland has officially unveiled Apertus, the country’s first national large language model (LLM) designed as a piece of public digital infrastructure. Unlike many commercial AI solutions, Apertus is open-source and built with transparency, compliance, and accessibility at its core. The project represents a bold step for Switzerland in ensuring technological sovereignty and setting a model for other European nations.
Developed through a collaboration between the École Polytechnique Fédérale de Lausanne (EPFL), ETH Zurich, and the Swiss National Supercomputing Centre (CSCS), Apertus has been released in two configurations: an 8-billion parameter model and a 70-billion parameter model. Access is available through Swisscom or via the Hugging Face platform, making it widely available for researchers, developers, and businesses.
One of the defining aspects of Apertus is the training methodology. The model was trained on 15 trillion tokens across over 1,000 languages, with 40% of the dataset representing non-English texts. It also includes Swiss national languages, such as Swiss German dialects and Romansh, which positions the model as uniquely adapted for local cultural and linguistic contexts. Importantly, the training relied only on open data sources, with crawlers respecting websites’ indexing rules. This level of data ethics sets Apertus apart from many proprietary AI systems that have faced criticism for questionable data sourcing practices and copyright concerns.
Transparency is central to Apertus. The developers released not just the model weights, but also the training code, documentation, and full dataset sources. For European businesses operating under strict GDPR and intellectual property laws, this openness offers an advantage: organizations can confidently use Apertus without the risk of hidden legal issues. Already, banks and financial institutions are exploring the model as a compliant, privacy-respecting AI solution, particularly for sectors where data security and confidentiality are critical.
Beyond Switzerland, the launch of Apertus illustrates a growing trend of state-backed AI initiatives. Governments increasingly recognize AI as strategic infrastructure, not merely a commercial tool. By creating open, national AI systems, countries aim to reduce dependence on private tech giants while maintaining control over data governance and regulatory compliance. For Europe, where digital sovereignty is a pressing issue, Apertus could serve as a blueprint for other nations to follow. If Switzerland’s initiative succeeds, we may soon see similar national LLM projects emerging across the European Union.
Conclusion: With the release of Apertus, Switzerland has positioned itself at the forefront of ethical, transparent, and sovereign AI development. By treating AI as public infrastructure rather than a proprietary asset, the country is paving the way for a more open, secure, and regulation-friendly future. Apertus is not just a model; it is a statement about how nations can reclaim control over their digital ecosystems.
