Switzerland has introduced Apertus, a groundbreaking, fully open-source large language model. This multilingual LLM was trained on an immense 15 trillion tokens and encompasses over 1,000 languages. A key differentiator for Apertus is its commitment to complete openness, granting access to model weights, training datasets, documentation, and intermediate checkpoints. Unlike proprietary models, Apertus' source code and all training materials are released under a permissive open-source license allowing commercial use. The detailed and reproducible training process enables researchers and watchdogs to audit data sources and verify compliance with laws. Apertus development strictly followed Swiss data protection and copyright laws, including retroactive opt-out mechanisms. From a privacy standpoint, the model exclusively uses publicly available data, meticulously filtered to exclude personal information and respect opt-out signals. This approach aligns with regulatory frameworks like the EU AI Act and showcases privacy-respecting AI. Developed for the public good, Apertus is envisioned as a public digital infrastructure. The model is accessible through Swisscom's Sovereign Swiss AI Platform, Hugging Face, and the Public AI Inference Utility.