Switzerland has officially unveiled its own national AI language model, Apertus LLM. It was developed by the country's leading universities, and the main idea is to create a transparent and inclusive alternative to ChatGPT, Llama, and others.
Apertus LLM is compared in power to Meta's Llama 3 (2024), although the Swiss emphasize that they are not trying to compete with American or Chinese giants, which invest billions every year in the development of newer models. Instead, the focus is on open source and security — each user can upload the model to their own server, while maintaining control over the data.
The model has already been called a "new phase" in the history of open LLMs. Unlike many competitors, Apertus discloses all technical details, including the "recipe" for learning, which should increase trust and reduce the risks of bias or "hallucinations."
Apertus is expected to find applications in scientific research, medicine, education and even finance, where privacy and compliance with European regulations are particularly important. The Swiss Bankers Association believes its own LLM has "great long-term potential," although banks, including UBS, are already working with OpenAI and Microsoft.
Despite business skepticism, the project received support from the state and the scientific community as a strategic step in preserving digital sovereignty.