DFM Mimir v1 is a 1B hierarchical-reasoning model trained only on permissible data, competing with Qwen 3.5 4B.

Danish Foundation Models trained it from scratch on 161 datasets. Across 20 benchmarks spanning English, math and code, and Danish, it beats the original HRM-Text 1B, sets a new Danish state of the art, and competes with Qwen 3.5 4B and Gemma 4 E2B. Weights are on Hugging Face. arXiv 2608.13517 The claim that matters: a licensing-clean data pipeline no longer costs an order of magnitude in capability at the small end.