Switzerland’s Apertus: A Landmark in the Open Source AI Revolution

In a significant move that signals the growing importance of transparency and accessibility in artificial intelligence development, Switzerland has released Apertus, a fully open AI model that stands as a powerful alternative to proprietary systems. Built by a consortium of leading Swiss institutions including EPFL, ETH Zurich, and the Swiss National Supercomputing Centre (CSCS), this groundbreaking initiative represents a fundamental shift in how AI technology can be developed and shared with the global community.

The name “Apertus” comes from the Latin word for “open,” reflecting the project’s core philosophy of complete transparency. Unlike many commercial AI systems that reveal only select technical details, Apertus makes its entire development process accessible to the public. This includes its architecture, training data, model weights, and comprehensive documentation, establishing a new standard for openness in advanced AI systems.

Available in two versions—an 8-billion-parameter model and a more powerful 70-billion-parameter variant—Apertus can be downloaded directly from Hugging Face or accessed through Swisscom, a strategic partner in the initiative. Both versions are released under a permissive open-source license, enabling their use across research, education, and commercial applications without the restrictive terms often associated with proprietary AI systems. For those looking to deepen their understanding of building with Hugging Face tools, Hugging Face in Action provides valuable insights into the platform that hosts Apertus.

“With this release, we aim to provide a blueprint for how a trustworthy, sovereign, and inclusive AI model can be developed,” said Martin Jaggi, Professor of Machine Learning at EPFL and member of the Steering Committee of the Swiss AI Initiative.

This sentiment reflects growing concerns about AI sovereignty, particularly in regions seeking technological independence from dominant global players. The comprehensive exploration of these sovereignty issues is well-documented in The Quest for AI Sovereignty, Transparency and Accountability.

What sets Apertus apart from many existing AI models is its remarkable multilingual capabilities. The training process involved an impressive 15 trillion tokens across more than 1,000 languages, with approximately 40% of the data being non-English. This deliberate focus on linguistic diversity enables Apertus to support languages often overlooked in mainstream AI development, including Swiss German and Romansh. This approach addresses a critical gap in current AI systems, which tend to perform significantly better in English than in other languages, perpetuating digital divides.

Thomas Schulthess, Director of CSCS and Professor at ETH Zurich, emphasized that Apertus represents “a driver of innovation and a means of strengthening AI expertise in research, society and industry.” The project’s vision extends beyond typical technology transfer models, instead positioning AI as a form of public infrastructure—similar to essential utilities like highways, water, or electricity.

The Swiss approach to AI development aligns closely with emerging regulatory frameworks, particularly the European Union’s AI Act. Apertus was developed with careful attention to Swiss data protection laws, Swiss copyright regulations, and the transparency requirements outlined in EU legislation. The dataset was deliberately limited to publicly available information, with personal data removed and website opt-out requests honored. Furthermore, ethical guidelines were applied to exclude inappropriate material before training began. For those navigating these complex regulations, Navigating the EU AI Act: The Annotated Regulation offers valuable guidance on compliance and implementation.

“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,” noted Imanol Schlag, technical lead of the project and Research Scientist at ETH Zurich.

While accessing and using Apertus is straightforward for experienced developers, practical implementation requires appropriate computational resources. To facilitate broader adoption, developers can test the model during the ongoing Swiss AI Weeks, which continue until October 5, 2025. Hackathon participants will gain access through Swisscom-hosted interfaces, while Swisscom business customers can already utilize the model through the company’s AI platform. For international users, Apertus will be available through the Public AI Inference Utility.

Switzerland’s entry into the open-source AI landscape comes at a time when the market dynamics for AI models are rapidly evolving. French AI company Mistral, for example, is reportedly close to securing a $14 billion valuation in its latest funding round, demonstrating the significant commercial interest in AI technologies. However, Switzerland’s approach with Apertus represents an alternative vision—one where AI is developed as a public good rather than primarily as a commercial product. Developers interested in exploring similar open-source AI implementations can find practical guidance in OPEN SOURCE LLMs: Customize and Deploy LLaMA, Mistral, and GPT4All with Real Projects.

This philosophy of openness in technology is gaining momentum in unexpected places. Recently, Microsoft made headlines by open-sourcing its historic 6502 BASIC interpreter from 1976, which was originally written by Bill Gates and early Microsoft employee Ric Weiland. This software played a crucial role in the early personal computer revolution, powering machines like the Apple II, Commodore PET, and Commodore 64. Microsoft’s decision to release this code under an MIT license after nearly 50 years reflects a broader shift toward open-source principles even among companies that historically favored proprietary models.

The future plans for Apertus are ambitious. The Swiss team intends to expand the model family, improve efficiency, and develop domain-specific tools for sectors including law, healthcare, climate science, and education. Throughout this evolution, they remain committed to maintaining rigorous standards of transparency and accessibility.

“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,” said Antoine Bosselut, Professor at EPFL and Co-Lead of the Swiss AI Initiative.

As nations and organizations worldwide grapple with questions of AI development, implementation, and regulation, Switzerland’s approach with Apertus offers a compelling alternative to closed commercial systems. By prioritizing transparency, accessibility, and public benefit, it demonstrates how advanced AI technologies can be developed in ways that empower broader communities rather than concentrating power in the hands of a few technology giants. The success of this model could significantly influence the future trajectory of AI development globally, potentially accelerating the shift toward more open, collaborative approaches to technological advancement.