DFM Mimir v1 Release

Danish Foundation Models has released Mimir v1, a 1-billion-parameter language model based on the Hierarchical Reasoning Model (HRM) architecture and trained from scratch on permissible and openly licensed data.

Mimir is optimised for Danish and English and was trained on a mixture of 161 datasets. Across 20 benchmarks covering English, mathematics and code, and Danish, it outperforms the original HRM-Text 1B and achieves particularly strong results for Danish.

The work originated within OdenseNLP and was developed as part of the Danish Foundation Models project.

The model is available on HuggingFace.

Read the full release note on the DFM Website and the technical report here.