Swiss innovation is making waves in the artificial intelligence space with the release of a groundbreaking open AI model named Apertus. Developed collaboratively by renowned institutions—including the École Polytechnique Fédérale de Lausanne (EPFL), ETH Zurich, and the Swiss National Supercomputing Centre (CSCS)—Apertus stands out for its commitment to accessibility and transparency. The name “Apertus,” which translates to “open” in Latin, encapsulates the model’s core philosophy: complete openness in its design and training process.
Designed to empower developers and organizations alike, Apertus can be utilized for various applications, from chatbots and translation tools to education-focused programs. Users have the flexibility to download the model directly from platforms like Hugging Face or access it through Swisscom, a key strategic partner. Two distinct versions are available—a compact 8-billion-parameter model and a robust 70-billion-parameter version—both licensed under a permissive open-source agreement that facilitates use in research, education, and commercial settings.
Built for Openness
What sets Apertus apart from other AI models is its unparalleled transparency. Unlike many proprietary systems that withhold crucial details, Apertus offers a complete view of its architecture, training data, and extensive documentation. As per Martin Jaggi, Professor of Machine Learning at EPFL and a member of the Swiss AI Initiative’s Steering Committee, this initiative aspires to establish a blueprint for creating a reliable, sovereign, and inclusive AI model. Regular updates will be driven by a dedicated team of engineers and researchers from CSCS, ETH Zurich, and EPFL.
Thomas Schulthess, Director of CSCS and Professor at ETH Zurich, emphasized that Apertus isn’t your typical technology transfer from research to product; instead, it’s a foundational infrastructure initiative meant for long-term impact and innovation enhancement across research, society, and industry.
Multilingual Reach
Training Apertus involved an impressive 15 trillion tokens derived from over 1,000 languages, where approximately 40% of the data consisted of non-English sources. Notably, it prioritizes languages frequently overlooked in large language models (LLMs), such as Swiss German and Romansh. This commitment to linguistic inclusivity is a testament to the model’s focus on the public good.
Imanol Schlag, the project’s technical lead and a Research Scientist at ETH Zurich, highlighted that Apertus ranks among the few fully open LLMs at this scale, intertwining multilingualism and transparency as its fundamental design tenets. Swisscom is already in the process of integrating Apertus into its sovereign AI platform. Daniel Dobos, Research Director at Swisscom, emphasized the initiative’s aim to foster a secure AI ecosystem that promotes public interest and enhances Switzerland’s digital sovereignty.
Testing the Open AI Model: Access and Real-World Use
For those eager to explore Apertus, the downloading process is straightforward for tech-savvy users; however, effectively deploying it often involves utilizing servers, cloud resources, or dedicated interfaces. Developers can begin testing Apertus during the ongoing Swiss AI Weeks, which will run until October 5, 2025. Participants in hackathons will have exclusive access through a Swisscom-hosted interface, while Swisscom business customers can implement the model immediately via the company’s AI platform. For international audiences, Apertus will also be available through the Public AI Inference Utility.
Joshua Tan, Lead Maintainer of the Public AI Inference Utility, asserted that Apertus exemplifies a leading public AI model, demonstrating that AI can function as a public infrastructure akin to highways and utilities.
Transparency and Compliance
The open-source license associated with Apertus ensures availability of the training data, model weights, and intermediate checkpoints for public scrutiny. The training methodology adhered strictly to Swiss data privacy regulations, copyright laws, and the EU AI Act’s transparency mandates. A significant focus was placed on ethical guidelines to filter out personal information and unwanted material prior to initiation of the training process.
The Future of Switzerland’s Open AI Model
Antoine Bosselut, Professor at EPFL and Co-Lead of the Swiss AI Initiative, maintained that Apertus underscores the possibility of generative AI being both potent and accessible. The release marks not just a final milestone, but the commencement of a long-term journey committed to fostering open, reliable, and sovereign AI structures for the benefit of the global community.
Looking ahead, there are ambitious plans for future updates that aim to expand the model family, boost efficiency, and develop domain-specific tools in critical areas, including law, health, climate, and education. This ongoing effort will continue to uphold rigorous standards of transparency, ensuring that Apertus remains at the forefront of AI development for public good.
(Photo by Cory Johnson)
See also: Microsoft gives free Copilot AI services to US government workers
Want to learn more about AI and big data from industry leaders? Explore the AI & Big Data Expo taking place in Amsterdam, California, and London. For comprehensive insights, it is part of TechEx and is co-located with other leading technology events. Click here for more information.
AI News is powered by TechForge Media. Discover other upcoming enterprise technology events and webinars here.
Inspired by: Source
