Apertus, the Swiss LLM: A New Step Toward Sovereign, Open, and Multilingual AI

In early September 2025, EPFL, ETH Zurich, and the Swiss National Supercomputing Centre (CSCS) unveiled Apertus, a groundbreaking large language model (LLM) that is fully open source and supports more than 1,000 languages, including Swiss regional languages like Romansh and Swiss German. This ambitious initiative marks a milestone for AI transparency, sovereignty, and inclusivity, offering Swiss SMEs and managers a unique opportunity to harness AI within a trustworthy and local framework.

What is Apertus?

Apertus—Latin for “open”—is distinct for its full transparency: its code, training data, and model weights are completely accessible under the Apache 2.0 license. Trained on 15 trillion tokens on Switzerland’s Alps supercomputer, it provides two models (8 billion and 70 billion parameters), enabling both individual and enterprise-scale uses.

Its multilingual design includes roughly 40% of non-English data, a deliberate choice to represent many lesser-supported languages and dialects vital to Swiss culture.

How does Apertus compare to other LLMs?

FeatureApertus (Switzerland)GPT-4 (OpenAI)LLaMA 2 (Meta)
Open sourceFully open (code, weights, data)Proprietary, restricted accessOpen source but partial components
Languages supported>1,000 including Swiss dialectsPrimarily EnglishMultilingual, English-focused
Training infrastructureSwiss supercomputing center (CSCS)Private cloud infrastructurePrivate clusters
LicenseApache 2.0 (permissive)Commercial licensesOpen source with restrictions
Transparency & complianceFull transparency, GDPR compliantLimited transparencyVariable transparency
CustomizabilityFully adaptable and inspectableSomewhat customizableCustomizable
Target usesLocal applications, SMEs, researchBroad commercialResearch & enterprise

Benefits for Swiss SMEs and Managers

  • Digital sovereignty: Keep sensitive data within Swiss jurisdiction, ensuring compliance and data security.
  • Inclusive multilingualism: Better serve customers and teams by supporting regional languages.
  • Innovation-friendly openness: Encourage co-development, research, and tailor-made AI solutions.
  • Ethical AI: Built on transparency and compliance aligning with Swiss and EU standards, fostering user trust.

Limitations and Outlook

Apertus is still evolving, with ongoing improvements planned in task specialization, inference speed, and fairness. The team behind it includes researchers from EPFL, ETH Zurich, and CSCS committed to continuous updates focusing on sectors like health, education, and law.

How Ixtria supports you

Ixtria stands ready to help Swiss SMEs navigate and adopt Apertus-based AI safely and effectively. From consulting and integration to tailored training, we help you leverage this new era of trustworthy and sovereign AI technology.


For further reading


Leave a Comment

Your email address will not be published. Required fields are marked *

Scroll to Top