Apertus 70B

Apertus 70B is a large language model developed through a collaboration between ETH Zurich, EPFL, and the Swiss National Supercomputing Centre (CSCS), and it is positioned as a truly open release. The model is hosted on Hugging Face under the swiss-ai organization, where the Apertus-70B-2509 release serves as the concrete open-weights artifact supporting that positioning. By making both weights and development intent publicly available, the project aims to give researchers and developers a foundation model whose behavior and lineage can be inspected and built upon rather than treated as a black box.

For practical use, Apertus 70B is aimed at teams that want an open-weights text model they can run, fine-tune, and audit on their own terms. Its release has drawn notable attention in the open-source community, with the announcement thread on Hacker News generating significant discussion around transparency and reproducibility practices. The model fits workflows where openness and provenance matter as much as raw capability, making it a reasonable choice for academic studies, internal assistants, and downstream fine-tuning experiments that benefit from a well-documented European-built base model.