Switzerland’s latest large language model, developed by ETH Zurich, EPFL, and CSCS, marks a significant step toward transparent, ethical AI. Trained on the Alps supercomputer using 100% carbon-neutral energy, the model will be fully open-source under an Apache 2.0 licence, with release expected summer 2025.
Available in two sizes: 8B and 70B parameters, AlpineGPT supports over 1,000 languages, with 40% of its training data drawn from non-English sources. This focus on linguistic diversity makes it one of the most inclusive LLMs to date. The model adheres to Swiss and EU data protection standards, ensuring auditability and reproducibility for researchers and public institutions.
Built as public infrastructure and not a commercial product, AlpineGPT sets a high bar for sustainable, multilingual AI innovation.
Learn more and follow release updates via ETH Zurich:
ethz.ch – A language model built for the public good
