Apertus: The New Open-Source AI from Switzerland

The AI world keeps moving – and now our neighbor Switzerland is joining the race in full force. Apertus, a fully open-source language model from Switzerland, has been making waves since September 2025. Developed by ETH Zurich, EPFL, and the Swiss National Supercomputing Centre in Lugano, Apertus aims to be more than just another alternative to ChatGPT & Co. – it stands for transparency, compliance, and European sovereignty.

But what makes Apertus so special? The model is completely open source: from parameter weights and training data to the algorithm itself, everything is publicly accessible. This makes it stand out from the major players in the U.S. and China (think DeepSeek). For many companies, the key question is: Can Apertus play a role in marketing, content creation, and internal applications?

In our guides on the best AI chatbots and GDPR-compliant ChatGPT alternatives, we’ve already introduced several privacy-focused solutions like Mistral – and now, with Apertus, the next exciting chapter in AI development is unfolding. The question remains: Can it keep up with the big players?

What To Expect

• What is Apertus?
• Why is Apertus exciting?
• Reactions & Criticism of Apertus
• Exciting Use Cases + Quickstart with Apertus
• What Does Apertus Mean for You?
• Conclusion: Time to Get Started with Apertus
• FAQ about Apertus

CONTACT

Apertus Setup with WEVENTURE

CALL DIRECTLY?

Book a call with our experts.

 

BOOK A CALL

CALL DIRECTLY

What is Apertus?

Apertus is a fully open, multilingual Large Language Model (LLM) from Switzerland – developed by EPFL, ETH Zurich, and the Swiss National Supercomputing Centre (CSCS) as part of the Swiss AI Initiative. The team has made the architecture, weights, training data, and training recipes publicly available and well-documented. The goal is to provide a trustworthy, transparent foundation that both researchers and companies can build upon. According to ETH Zurich, Apertus was officially released on September 2, 2025.

Key Features

• Two model sizes: 8 billion and 70 billion parameters – both openly available.
• Training & Languages: ~15 trillion tokens, covering over 1,000 languages (including Swiss dialects such as Swiss German and Romansh); roughly 40% of the data is non-English.
• Transparency & Compliance: Trained exclusively on publicly available data, respecting machine-readable opt-outs (including retroactively) and adhering to EU AI Act transparency requirements.
• Open-Source License: Permissive (Apache-2.0, according to the Model Card).
• Technical Specs: Decoder-only Transformer architecture, 65,536-token context window, with open technical report and checkpoints.
• Mission: “Built for the public interest” – with transparency, multilingualism, and compliance as core principles.

Access & Hosting

• Hugging Face: for downloads and model access.
• Public AI: for chat and inference.
• Swisscom’s sovereign Swiss-AI platform: business access within Switzerland.
• AWS SageMaker: available for enterprise deployment.

Why is Apertus exciting?

Apertus isn’t just another new language model. It stands out through three key aspects that make it particularly interesting for European companies and organizations:

1. Transparency & Compliance

Few other LLMs are as clearly documented in terms of what they were trained on. Its architecture, weights, data sources, and training methods are all public – a first at this scale. This makes Apertus fully compliant with the EU AI Act’s transparency requirements and sets it apart from U.S. and Chinese models, whose training data often remain opaque. For companies that value traceability and GDPR compliance, that’s a real competitive edge.

2. European Sovereignty

Until now, the market has been dominated by U.S. models like ChatGPT, Gemini, Grok, and Perplexity, as well as Chinese projects like DeepSeek. With Apertus, Switzerland is sending a strong signal: high-performance AI can be built in Europe – independently of Big Tech. This strengthens digital sovereignty and opens new paths for national initiatives and research partnerships.

3. Linguistic Diversity

Supporting over 1,000 languages, including Swiss dialects like Romansh, Apertus demonstrates a range few existing models can match. For internationally operating companies developing multilingual content, this is a major advantage.

Kurzer Vergleich: Apertus, Mistral & ChatGPT

• Mistral: Our current standard for GDPR-compliant AI applications – high-performing, reliable, and privacy-safe.
• ChatGPT: Excellent in usability and features, but not suitable for sensitive data. We use ChatGPT for most SEO-related content.
• Apertus: The new open-source project with maximum transparency and a strong focus on compliance – currently less mature, but with enormous potential.

If you’d like to dive deeper: In our guide on GDPR-compliant ChatGPT alternatives, we explain when each solution makes sense. And in our AI chatbot overview, you’ll find specific use cases for projects that don’t involve sensitive data.

Reactions & Criticism of Apertus

Since its launch in early September 2025, Apertus has generated significant buzz within the AI community – and opinions couldn’t be more divided.

On one hand, there’s great interest and positive feedback: According to ETH researcher and co-developer Imanol Schlág, the model has already been downloaded more than 350,000 times. Encouragingly, other national initiatives and research teams around the world are planning to build upon Apertus. The fact that a European project is even being mentioned in the same breath as Big Tech, Schlág says, is a success in itself – especially given the far smaller resources involved.

On the other hand, some critics argue that Apertus still can’t match the performance of ChatGPT. Schlág’s response: “Finding flaws is easy. But the fact that we’re even being compared to Big Tech is actually fantastic.”

A closer look shows that the smaller 8-billion-parameter model already outperforms others of similar size. The larger 70B version is more powerful but still needs refinement before it can be considered a fully mature product.

What sets Apertus apart most clearly, according to the developers, is its unique commitment to transparency: it is the largest model that meets the EU AI Act’s compliance requirements. This makes it stand out sharply from the proprietary models coming from the U.S. and China.

As for what’s next, Schlág says the focus will now shift toward multimodality – enabling the model to understand images and audio. This could unlock new possibilities in education, healthcare, and accessible communication. In parallel, the core team is expanding to ensure long-term development and stability.

Exciting Use Cases + Quickstart with Apertus

Apertus isn’t a finished mass-market product like ChatGPT – and that’s precisely its strength. It’s an open platform that can be flexibly adapted for research, businesses, and content projects.

Potential Use Cases for Apertus

• Content Workflows & SEO: Generate first drafts for blog articles, meta descriptions, or snippet ideas that can later be refined editorially.
• Internal Knowledge Chatbots: Company documents and policies can be structured so employees can securely access relevant information.
• Multilingual Applications: With support for 1,000+ languages (including Swiss dialects such as Romansh), Apertus is ideal for international content strategies and markets.
• Research, Education & Public Sector: Thanks to its transparent training data, Apertus is particularly appealing for public institutions that prioritize traceability and compliance.

Quickstart: How to Try Apertus

1. Choose Your Access Method:
   • Via Hugging Face or Public AI (direct testing, including the 70B model, after login)
   • Through Swisscom’s sovereign Swiss-AI platform for business setups
   • Or via AWS SageMaker, where Apertus is also available
2. Select a Version:
   • The lighter 8B model for quick tests
   • The more powerful 70B model for advanced scenarios
3. Start Deployment:
   • Use a managed solution or set up self-hosting via standard inference servers.
4. Ensure Data Security:
   • For sensitive data, continue using Mistral.
   • For non-critical content, Apertus (or ChatGPT) can be safely applied.
5. Set Up a Pilot Project:
   • Begin with a clearly defined use case, measure results (e.g., content quality, efficiency gains), and scale gradually from there.

What Does Apertus Mean for You?

For your company, the key question isn’t just which AI tools exist — it’s which setup truly fits your needs. That’s exactly where we at WEVENTURE Performance come in:

• Sensitive data? → Our go-to recommendation remains Mistral. It’s proven, high-performing, and GDPR-compliant — ideal for confidential information or critical business processes.
• Efficiency & creativity? → For non-sensitive tasks like brainstorming, drafting, or quick responses, ChatGPT and Apertus are both valuable tools. Apertus also offers transparency and compliance, making it especially appealing for organizations committed to long-term European standards.
• Hybrid approach: In practice, we often combine multiple models depending on the use case. This way, you can benefit from each system’s strengths and design your workflows with flexibility.

At present, Mistral is our standard solution for privacy-compliant AI applications.

However, if Apertus delivers on its promises in real-world use, it has the potential to become the new standard itself. Until then, we’re actively testing it within our own content marketing processes — and can set up a customized environment for you that’s reliable today while leaving room for innovation tomorrow.

Conclusion: Time to Get Started with Apertus

With Apertus, Switzerland proves that AI development in Europe can be transparent, privacy-compliant, and innovative. The project is still in its early stages, but it’s already clear that its commitment to openness and compliance could make Apertus a serious contender against the big players in the medium term.

For companies, this means: stay alert. For sensitive data, we continue to rely on Mistral as our proven, GDPR-compliant solution. At the same time, we at WEVENTURE Performance are testing Apertus in our own content marketing workflows — to see where it truly adds value and how it fits into hybrid setups with Mistral and ChatGPT. So far, our impression is positive: while the model still makes a few mistakes, its output speed — similar to Mistral — is noticeably faster than ChatGPT. In the long run, we believe Apertus could evolve into a powerful alternative.

👉 Curious which AI setup makes the most sense for your business?
Let’s talk. We’ll provide tailored guidance, set up the right tools for you, and help you integrate AI into your daily operations — securely and efficiently.

FAQ about Apertus