This article presents the development of an ethical LLM co-created by researchers from EPFL, ETH Zurich, and CSCS. The LLM is a proof of concept exploring how large language models can still perform well without relying on massive data acquisition and traditional scaling approaches.