Paper Overview
- Field: NLP
- Authors: Peter Schneider-Kamp, Jacob Nielsen, Gianluca Barmina, Kenneth Enevoldsen, Lukas Galke Poech
- Published: 2026-08-13
- arXiv: 2608.13517
- 1B-parameter model based on the HRM architecture, trained from scratch
- Uses only permissible post-training data, lowering barriers for ethically sourced LLM development
- Trained on a mixture of 161 datasets
- Outperforms the original HRM-Text 1B
- Competitive with larger frontier models (Qwen 3.5 4B, Gemma 4 E2B) across 20 benchmarks (English, Math & Code, Danish)
- Sets a new state of the art for Danish
Abstract
Current large language model development relies on massive, often non-permissible datasets, creating a high barrier for researchers committed to open-source and ethically sourced data. The authors introduce Mimir v1, a 1-billion-parameter language model based on the Hierarchical Reasoning Model (HRM) architecture, that is trained from scratch and delivers highly competitive performance for English and sets a new state of the art for Danish using only permissible post-training data.Trained on a mixture of 161 datasets, Mimir v1 outperforms the original HRM-Text 1B and competes with larger frontier models like Qwen 3.5 4B and Gemma 4 E2B, tested across 20 benchmarks for English, Math & Code and Danish.
Key Highlights
Model Availability
The model is available on the Hugging Face Hub: https://huggingface.co/danish-foundation-models/DFM-Mimir--- *Automatically collected on 2026-08-15*