Switzerland launches transparent ChatGPT alternative

Apertus LLM is intended as a trustworthy and inclusive AI Large Language Model Keystone / Ennio Leanza

Switzerland has entered the artificial intelligence (AI) race with the launch of a national Large Language Model (LLM) to provide an alternative to the likes of ChatGPT, Llama and DeepSeek.

+Get the most important news from Switzerland in your inbox

Designed by leading Swiss universities, the Apertus LLM is comparable to Meta’s Llama 3 model from 2024. Since then, Meta and other rivals have produced more advanced versions.

But the Swiss team insist they are not trying to compete with the multi-billion dollar budgets of United States trailblazers. They are happy to sacrifice the latest frills aimed at general users in favour of a safer and more accessible AI system for scientific researchers and commerce.

+ How Switzerland strives to steer a safe AI course

“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 the Swiss Federal Institute of Technology Lausanne EPFL.

Hallucinations and bias

The frenetic pace of AI innovation since the public release of OpenAI’s ChatGPT in 2022 has seen leading companies, such as Anthropic, release a spate of LLMs every year. Outside the US, China’s DeepSeek and Qwen platforms and models from France’s Mistral have deepened the pool of options for AI users.

But the race for AI dominance has also left a trail of problems: machines that hallucinate or magnify human bias and a growing list of copyright infringement lawsuits over the material that companies use to train their LLMs.

Artificial intelligence inspires both hopes of a better future and fears of inflicting societal damage if left to roam wild without controls.

One answer is to produce publicly available, open-source AI systems to slug it out for market share with privately owned rivals.

+ Swiss Alps supercomputer to leverage AI for science

Commercial LLMs tend to push the frontiers of innovation by releasing the most advanced platforms. Open-source models are free to use and allow users to inspect their designs to figure out how they generate responses.

Some models, such as DeepSeek from China, shine a light on how they operate but not how they were programmed in the first place. It’s like watching a chef prepare a dish without knowing what went into the sauce.

Open-source milestone

Apertus (from the Latin word meaning “open”) leaves nothing to the imagination in this respect. It promises that every nut and bolt is open for public scrutiny along with its design manual and recipe formula. This is intended to inspire public trust and assuage concerns about the potential ugly side of AI.

Apertus is one of the most ambitious open-source models to date, according to Leandro von Werra, head of research at the open-source AI community Hugging Face.

“It’s not totally unique, but it’s still a trailblazer given its scale and the amount of compute used to train the model,” he told Swissinfo. “It’s definitely a new milestone in open models.”

Researchers, programmers, start-ups and the public sector can download a copy of Apertus onto their own servers to build projects. This approach allows users to keep control over their data.

The scale and success of applications built on top of Apertus will ultimately reveal the true worth of the Swiss LLM. Scientists are already exploring projects in the fields of health, education and the climate.

Swiss industry groups have welcomed the home-grown AI initiative, particularly its focus on data security. But they caution it will face stiff competition for commercial attention from a growing range of powerful international rivals.

Competing for users

The Swiss Bankers Association believes a home-grown LLM has “great long-term potential” for the financial industry, especially given the need to comply with local data protection and banking secrecy laws.

However, some Swiss banks are already building AI projects using other LLMs. For example, Switzerland’s largest bank, UBS, is already working with OpenAI and Microsoft.

Swissmem, which represents the domestic electrical engineering and machine building industries, believes that a Swiss LLM could better serve local companies as it is built to respect European data regulations. But this is no guarantee that Swiss industry will jump onto Apertus en masse.

Potential users will gauge the speed, accuracy and depth of responses generated by Apertus compared to other LLMs. These results will be weighed up against the advantages of the system’s transparency and their data security requirements.

“Experience shows that there is no single solution that fits all needs,” Swissmem’s head of digitalisation, innovation and technology, Adam Gontarz, told Swissinfo.

“Each project must take into account the specific circumstances and requirements of the business,” he added. “In some cases, international solutions may also be the most effective choice.”

Edited by Gabe Bullard/vdv/ts