Switzerland Releases Its Own Open-Source AI Model "Apertus"

Switzerland has released its own open-source AI model called "Apertus." The model, developed with the involvement of Swiss research institutions, positions Switzerland as a serious player in the European AI ecosystem.

How It Came About

Behind Apertus stands the Swiss AI Initiative – a consortium of Swiss universities, research institutions, and public bodies. The goal was clear from the start: a language model that genuinely reflects the linguistic and cultural reality of Switzerland.

Existing models like GPT-4 or Mistral were primarily trained on English-language data. German, French, and Italian have reasonable coverage by now – but Swiss German, Romansh, or an understanding of Swiss legal terminology and administrative language? That is barely present in commercial models. Then there is the data protection question: who guarantees that inputs to US-based AI services are not used for training?

What Is Apertus?

Apertus is a large language model (LLM) developed specifically for the Swiss context. Its key features:

• Multilingualism: German, French, Italian, and Romansh – all four national languages
• Swiss German: Understanding of dialect variants
• Open Source: The code is publicly accessible, auditable, and customizable
• Data protection: Trained with a privacy-by-design approach, compliant with the Swiss Data Protection Act (nDSG)

Significance for Swiss Companies

Apertus opens new possibilities for companies that previously hesitated to use US-based AI services:

Data sovereignty: The model can be operated fully locally – no data leaves the company or Switzerland. For firms working with sensitive customer, patient, or financial data, this is a significant advantage over cloud-based services.

Compliance: For regulated industries (finance, healthcare, public administration), a Swiss model offers clear advantages. No data transfer to the US, no dependency on Privacy Shield agreements or CLOUD Act risks.

Adaptability: Open source means the model can be fine-tuned on your own data, processes, and specialized terminology – something that is barely possible with proprietary models.

Swiss language competence: Companies serving customers across all linguistic regions, or communicating with Swiss authorities and partners, benefit from a model that genuinely understands local language – not just translates it.

Limitations

Apertus is a promising step, but not a complete replacement for established models like GPT-4 or Claude. For many use cases – complex analysis, code generation, multimodal tasks – commercial models remain more capable.

Apertus's strength lies specifically in Swiss language competence and the ability to operate locally. As a complement to an existing AI strategy, it is interesting for many Swiss companies – not as a standalone solution.

Conclusion

The release of Apertus is an important signal: Switzerland takes AI sovereignty seriously. For companies seeking data protection-compliant AI solutions – particularly where Swiss language competence, local deployment, or compliance requirements play a role – it is worth a closer look. For an overview of other leading AI assistants, see our article on Claude by Anthropic.

Whether Apertus is right for your specific use case is something we are happy to explore together in an appointment.