With Soofi S the Soofi consortium presented the first building block of a European AI model family. The project aims to develop high-performance open-source AI base models on European infrastructure and provide companies, research institutions, start-ups, and public administration with a transparent and sovereign alternative to non-European language models.

Soofi S was specifically developed for industrial applications. The model is suitable for analyzing technical and regulatory documents, code generation, as well as agent-based AI applications. Thanks to a modern Mamba Transformer architecture, the base model combines high performance with comparatively low energy consumption and, according to the consortium, achieves top scores among open AI models for German-language applications.

A special feature of the project is consistent transparency. In addition to model weights, the consortium also publishes technical documentation on training methods and data pipelines. This allows companies to more easily review, adapt the model to their own requirements, and operate it on sovereign infrastructure.

In the project funded by the Federal Ministry for Economic Affairs and Energy under the IPCEI-CIS-/8ra initiative, numerous German research institutions and AI companies are involved, including Fraunhofer IAIS, Fraunhofer IIS, DFKI, hessian.AI, CAIDAS at the University of Würzburg, as well as the L3S Research Center at Leibniz University Hannover. The consortium is coordinated by the AI Association.

Companies can now participate as pilot users. The consortium tests Soofi S together with industry partners in practical application scenarios and seeks further companies wanting to test and jointly develop the model in real processes. Interested parties can contact the project team through contact@soofi.info the project team.

Companies looking to integrate Soofi S or other generative AI solutions into their operations can use the free BAIOSPHERE AI Compass as a practical guide. It helps organizations navigate key aspects of AI adoption, from developing an AI strategy and assessing data and infrastructure requirements to identifying funding opportunities and connecting with the right partners in the AI ecosystem.