Apertus releases open large language model

A new large language model (LLM) called Apertus, meaning “open” in Latin, has been launched with a distinctive feature: the entire development process is openly accessible and fully documented. This includes the architecture, model weights, and the training data and recipes.

The LLM is designed to serve as a building block for developers and organisations creating applications such as chatbots, translation systems, and educational tools. It was developed through a joint initiative announced in July by the École Polytechnique Fédérale de Lausanne (EPFL), the Swiss Federal Institute of Technology in Zurich (ETH Zurich), and the Swiss National Supercomputing Centre (CSCS).

AI researchers, professionals, and experienced enthusiasts can access the model through strategic partner Swisscom or download it from Hugging Face, a platform for AI models and applications.

Apertus is available for free in two sizes, featuring 8-billion and 70-billion parameters. The smaller model is more suitable for individual use. Both versions are released under a permissive open-source license, enabling use in education, research, and a wide range of societal and commercial applications.

A fully open-source LLM

As a fully open language model, Apertus enables users to build upon the model and adapt it to their specific needs. Users can inspect any part of the training process. This distinguishes Apertus from models that make only selected components accessible.

“With this release, we aim to provide a blueprint for how a trustworthy, sovereign, and inclusive AI model can be developed,” says Martin Jaggi, professor of machine learning at EPFL and member of the steering committee of the Swiss AI Initiative.

The model will be regularly updated by the development team which includes specialised engineers and a large number of researchers from CSCS, ETH Zurich and EPFL.

A driver of innovation

With its open approach, EPFL, ETH Zurich and CSCS are venturing into new territory.

Thomas Schulthess, CSCS director and ETH Zurich professor, says: “Apertus is not a conventional case of technology transfer from research to product. Instead, we see it as a driver of innovation and a means of strengthening AI expertise across research, society and industry.”

EPFL, ETH Zurich and CSCS aim to provide foundational technology and infrastructure to foster innovation across the economy.

Apertus was trained on 15-trillion tokens covering more than 1,000 languages, with 40% of the data in non-English languages. It includes several that are underrepresented in large language models, such as Swiss German and Romansh.

Imanol Schlag, technical lead of the LLM project and research scientist at ETH Zurich, says: “Apertus is built for the public good. It stands among the few fully open LLMs at this scale and is the first of its kind to embody multilingualism, transparency, and compliance as foundational design principles.”

Daniel Dobos, Swisscom research director, says: “Swisscom is proud to be among the first to deploy this pioneering large language model on our sovereign Swiss AI Platform. As a strategic partner of the Swiss AI Initiative, we are supporting the access of Apertus during the Swiss {ai} Weeks.

“This underscores our commitment to shaping a secure and responsible AI ecosystem that serves the public interest and strengthens Switzerland’s digital sovereignty.”

Accessibility

Apertus practical use requires additional components such as servers, cloud infrastructure, or dedicated user interfaces. The upcoming Swiss {ai} Weeks will provide developers with the first opportunity to work directly with Apertus, test its capabilities, and offer feedback for future versions.

Swisscom will provide a dedicated interface for hackathon participants to facilitate interaction with the model. Swisscom business customers can access the Apertus model through the company’s sovereign Swiss AI platform.

For those outside of Switzerland, the external pagePublic AI Inference Utility will make Apertus accessible as part of a global movement for public AI.

Joshua Tan, lead maintainer of the Public AI Inference Utility, says: “Currently, Apertus is the leading public AI model: a model built by public institutions, for the public interest. It is our best proof yet that AI can be a form of public infrastructure like highways, water, or electricity.”

Transparency and compliance

Apertus has been developed with transparency as a central principle to ensure reproducibility of the training process. In addition to the models, the research team has released supporting resources, including documentation, source code of the training process and datasets, and model weights with intermediate checkpoints. All materials are available under a permissive open-source license that permits commercial use, with terms and conditions provided via Hugging Face.

The model was created in compliance with Swiss data protection and copyright laws, as well as the transparency requirements of the EU AI Act. The training corpus is limited to publicly available data, filtered to honour machine-readable opt-out requests from websites, including those made retroactively, and to remove personal data and other undesired content prior to training.

The beginning of a journey

Antoine Bosselut, professor and head of the Natural Language Processing Laboratory at EPFL and co-lead of the Swiss AI Initiative, says: “Apertus demonstrates that generative AI can be both powerful and open. The release of Apertus is not a final step, rather it’s the beginning of a journey, a long-term commitment to open, trustworthy, and sovereign AI foundations, for the public good worldwide.

“We are excited to see developers engage with the model at the Swiss {ai} Weeks hackathons. Their creativity and feedback will help us to improve future generations of the model.”

Future versions aim to expand the model family, improve efficiency, and explore domain-specific adaptations in fields like law, climate, health and education. Apertus says it expects to integrate additional capabilities, while maintaining strong standards for transparency.