A team of Swiss institutions has launched Apertus, a latest large language version (LLM) created to enhance research, education, and industrial applications. Established via EPFL, ETH Zurich, and the Swiss National Supercomputing Centre (CSCS), Apertus is absolutely open, with architecture, training data, and documentation publicly provided.
The model comes in two variations—an 8-billion parameter and a 70-billion parameter version—both certified for open use in research, education, and commercial projects. Apertus may be accessed on Hugging Face or through Swisscom, that is deploying it on its sovereign AI platform.
A Blueprint for tested AI
“With this launch, our goal is to give a blueprint for how a tested, sovereign, and inclusive AI model can be created,” stated Martin Jaggi, Professor of Machine Learning at EPFL. Regular updates might be managed by engineers and researchers from CSCS, ETH Zurich, and EPFL.
Thomas Schulthess, Director of CSCS, added that Apertus is “a driver of innovation and a means of strengthening AI expertise in research, society and industry.” He highlighted that the ventures objectives is to generate a long-term infrastructure, instead of a one-time technology transfer.
Multilingual and Inclusive Design
Apertus was trained on 15 trillion tokens across 1,000+ languages, with 40% of its dataset in non-English content. It consist of underrepresented languages including Swiss German and Romansh, making it one of the most inclusive LLMs to date.
“Apertus is generated for the public good,” stated Imanol Schlag, technical lead and Research Scientist at ETH Zurich. “It is the main of its kind to encompass multilingualism, transparency, and compliance as foundational layout concepts.”
Real-World Deployment and Access
Developers could be in a able to test Apertus during Swiss (ai) Weeks, running until October 5, 2025, with access offered by a Swisscom-hosted interface. Business customers of Swisscom can already use the model, at the same time as international users will advantage access through the Public AI Inference Utility.
“Recently, Apertus is the main public AI model: a model generated via public institutions, for the general public interest,” stated Joshua Tan, Lead Maintainer of the Public AI Inference Utility. “It is our best proof yet that AI can be a form of public infrastructure like highways, water, or electricity.”
Commitment to Transparency and Compliance
The launch follows strict data protection standards, Swiss copyright law, and the EU AI Act’s transparency rules. The training dataset included only of publicly available data, processed to prevent personal data and unwanted content.
Antoine Bosselut, Professor at EPFL and Co-Lead of the Swiss AI Initiative, explained the launch as “the starting of a journey, an long-term commitment to open, honest, and sovereign AI foundations.” Intended updates consist domain-specific variations for law, fitness, climate, and education.
