Apertus: What you need to know about the Swiss AI model

In the world of artificial intelligence, superlatives abound. Almost every week, we hear about new, even more powerful models from major international tech companies. Amid all this news, a development from Switzerland is drawing attention: the AI language model Apertus.

Naturally, this news has raised a few questions for us. What can Apertus really do? How does it compare to other models? And most importantly: When is it a viable option for a project?

The Core of Apertus: Transparency as a Principle

Apertus was founded by leading Swiss research institutes (ETH Zurich, EPFL, CSCS). Its most important feature is not its performance, but its commitment to complete openness. The name says it all: "Apertus" is Latin for "open."

But what exactly does that mean? Many well-known models, including those in the open-source community, are only “open weight.” That means you can download and use the trained model. Apertus goes a decisive step further: The training data, methods, and code are also fully transparent and documented. So you not only know, that not only what the model can do, but also, whereby he acquired his knowledge.

This emphasis on transparency and traceable origins is anything but a given in the world of AI. Added to this is a clear focus on multilingualism, which also takes into account Swiss linguistic peculiarities such as Romansh and Swiss German.

The Issue of Data Protection

One of the most pressing issues regarding the use of AI is data security. The technical solution here is the ability to run an AI model on one’s own servers or in a Swiss data center.

We need to be clear on this: Other robust open-source models, such as the Gemma Models by Google, DeepSeek Models or gpt-oss by OpenAI. So you don't have to rely on Apertus to run a self-hosted AI solution that complies with data privacy regulations.

The key difference with Apertus lies in the trust placed in the model's origin, as a look at the technical report reveals. Hidden within it are some decisions made by the developers that go far beyond the usual legal and ethical standards.

1. Data Protection with Foresight: The "Hindsight" Approach
   One detail stands out in particular and sets Apertus apart from almost all other open-source models: how it handles training data. The developers have not only respected website operators’ opt-out requests (via robots.txt) going forward, but have also applied them retroactively to all data archives dating back to 2013. Simply put: If a website operator decides today that her content should not be used for AI training, her data has been removed from the entire training process. This “retroactive respect” significantly minimizes legal risks.
2. Built-in Forgetfulness: Protection Against Copyright Blunders
   Many companies are concerned that an AI might inadvertently disclose copyrighted or private data that it has "memorized" during training. Apertus actively counteracts this: Using a special training method, referred to as the “Goldfish Objective” in the report, the model was specifically prevented from memorizing text passages word for word. The tests confirm that the risk of such unintentional reproduction of training data is significantly reduced. This is not merely a hope, but a built-in design feature for greater security.
3. The 'Swiss AI Charter': Clear Values
   At Apertus, the focus on “Swiss values” is more than just a marketing slogan. The report reveals the development of a “Swiss AI Charter”—a set of guidelines directly based on Swiss constitutional values such as neutrality, consensus-building, federalism, and data protection. This charter was even validated by an overwhelming majority (an average of 97.3% approval) in a survey of Swiss citizens. It serves as a guide for the model when dealing with difficult or controversial issues.

The choice of Apertus is therefore less a technical decision than a strategic one: the company is opting for a model whose development process is guided by local values such as transparency and legal compliance.

An honest assessment of performance: Solid, but not at the top

How does Apertus fare in a direct performance comparison? The benchmarks from the technical report confirm our initial assessment and paint a clear picture:

• Compared to proprietary models Compared to the latest models from OpenAI, Anthropic, or Google, there is still a performance gap, especially when it comes to complex logical inferences.
• In the field of open-source models as well Apertus currently ranks in the solid middle of the pack. Models of a similar scale sometimes demonstrate higher performance in global benchmarks for mathematics or programming.

However, the report also highlights specific strengths: In tests that cultural and regional knowledge When it comes to querying data, Apertus performs exceptionally well. This confirms that the focus on multilingual and European data is paying off.

An Important Foundation

Apertus is not a "ChatGPT killer" from Switzerland. However, it is a solid and trustworthy foundation on which to build. It proves that Switzerland has the expertise to develop its own, independent AI technology.

We see our role as finding the best solution for your specific needs. In projects where maximum performance or low operating costs are critical, we may still recommend other models.

However, given the high standards for data protection, transparency, and legal certainty, Apertus offers a great new option for using AI without many of the concerns that have existed in the past.

Haven't tried Apertus yourself yet? Go to https://publicai.co/ Apertus is publicly available and ready for you to test.

Update July 2026:

It was published in July 2026 Apertus 1.5 - Still with 8 and 70 billion parameters, but with significantly more capabilities.

Apertus now understands images and audio, offers a reasoning mode, and processes contexts four times longer, with approximately 262,000 tokens. Instruction following and tool calling have also been improved.

With Apertus Mini There are also models with 0.5, 1.5, and 4 billion parameters that can run locally on significantly less powerful hardware.

At the same time, a genuine Swiss ecosystem is taking shape around Apertus: The Swiss AI Initiative continues to advance open research through Public AI Apertus is available as a chat platform and API, and now offers several Swiss and international providers production-ready hosting and inference for Apertus. Initiatives such as the Swiss AI Weeks They also take the model from the lab and apply it to practical projects.

The conclusion remains: Apertus isn't a Swiss ChatGPT killer—and it doesn't have to be. Its strengths lie in openness, transparency, and digital sovereignty. But now, you have to make far fewer compromises to achieve that.