Apertus: The Groundbreaking Open AI Model for Public Good
Apertus, a revolutionary open AI model developed by a coalition of Swiss institutions—including the École Polytechnique Fédérale de Lausanne (EPFL), ETH Zurich, and the Swiss National Supercomputing Centre (CSCS)—is set to redefine the landscape of artificial intelligence. Designed with transparency and accessibility at its core, Apertus serves as a foundation for future research and applications, offering developers a robust tool for various innovations.
What is Apertus?
The name “Apertus,” derived from Latin meaning “open,” aptly reflects the model’s commitment to openness. Unlike many proprietary AI systems, Apertus provides access to its entire architecture, training data, and documentation, allowing for inspection and further development by the public. This model can be utilized to create chatbots, translation tools, and educational applications.
How to Access Apertus
Developers and organizations interested in leveraging Apertus can download it directly from Hugging Face or access it via Swisscom, a strategic partner in the initiative. Apertus is available in two versions: an 8-billion-parameter model and a larger 70-billion-parameter variant, both released under a permissive open-source license. This means they can be used for research, education, and commercial projects without significant restrictions.
Built for Openness
Martin Jaggi, Professor of Machine Learning at EPFL and a member of the Steering Committee of the Swiss AI Initiative, emphasizes the importance of this release by stating, “With this release, we aim to provide a blueprint for how a trustworthy, sovereign, and inclusive AI model can be developed.” Apertus will be continuously updated by a dedicated team of engineers and researchers from CSCS, ETH Zurich, and EPFL.
According to Thomas Schulthess, Director of CSCS and Professor at ETH Zurich, Apertus is “a driver of innovation and a means of strengthening AI expertise in research, society, and industry.” This project is more than just a technology transfer; it represents a long-term infrastructure investment for the future.
Multilingual Capabilities
Apertus is not just another language model; it is built to accommodate a rich tapestry of languages. The training process utilized 15 trillion tokens across more than 1,000 languages, with 40% of the data sourced from non-English languages. This includes languages often overlooked by large language models (LLMs), such as Swiss German and Romansh.
Imanol Schlag, the technical lead of the project and Research Scientist at ETH Zurich, states, “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.”
Real-World Applications and Testing
While downloading Apertus is straightforward for experienced users, practical applications require adequate server capacity, cloud resources, or dedicated user interfaces. Developers can start testing Apertus during the Swiss AI Weeks, which run until October 5, 2025. Additionally, hackathon participants will gain access through a Swisscom-hosted interface, while Swisscom business customers can utilize the model on the company’s AI platform starting today. For international users, Apertus will be accessible via the Public AI Inference Utility.
Joshua Tan, Lead Maintainer of the Public AI Inference Utility, underscores the significance of this initiative: “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 similar to highways, water, or electricity.”
Transparency and Compliance
One of the standout features of Apertus is its adherence to transparency and legal compliance. Under its open-source license, training data, model weights, and intermediate checkpoints are readily available for public scrutiny. The training process strictly followed Swiss data protection regulations, Swiss copyright laws, and the transparency requirements outlined in the EU AI Act.
The dataset was meticulously curated to include only publicly available information, excluding any personal data and respecting website opt-out requests. Ethical guidelines were also implemented to filter out inappropriate content prior to commencing the training process.
A Future Committed to Open AI
Antoine Bosselut, Professor at EPFL and Co-Lead of the Swiss AI Initiative, states, “Apertus demonstrates that generative AI can be both powerful and open.” He emphasizes that the release of Apertus is merely the beginning of a journey, emphasizing a long-term commitment to establishing open, trustworthy, and sovereign AI foundations for the public good worldwide.
Future updates for Apertus aim to expand the model family, enhance efficiency, and develop domain-specific tools tailored for sectors such as law, health, climate, and education—all while maintaining strict standards of transparency and ethical responsibility.
Conclusion
The launch of Apertus marks a significant milestone in the evolution of open AI models. By prioritizing accessibility, transparency, and ethical compliance, Apertus sets a new standard for future developments in artificial intelligence. As we move forward, the implications for research, education, and industry are profound, paving the way for a more inclusive and responsible AI ecosystem.
FAQs
1. What is Apertus, and who developed it?
Apertus is an open AI model developed by Swiss institutions including EPFL, ETH Zurich, and CSCS, designed to serve as a foundation for research and applications.
2. How can I access the Apertus model?
You can download Apertus directly from Hugging Face or access it through Swisscom’s AI platform.
3. What makes Apertus different from other AI models?
Apertus is fully open, providing access to its architecture, training data, and documentation, unlike many proprietary AI systems.
4. What languages does Apertus support?
Apertus supports over 1,000 languages, with approximately 40% of its data sourced from non-English languages, including Swiss German and Romansh.
5. How does Apertus ensure ethical compliance?
The model adheres to Swiss data protection regulations and EU AI Act transparency requirements, filtering out personal data and unwanted material during its training process.
