DFM-Mimir

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. We introduce Mimir, 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, comprising approximately 70.479 billion tokens per epoch. Mimir 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.

Evaluation

Mimir is evaluated across Danish, English, and Math & Code benchmarks. The figure below shows average performance by subject area across compared models.

English benchmark results (Best scores in bold.)

English benchmark results (full datasets).

Math & Code benchmark results (Best scores in bold.)

Danish benchmark results (Best scores in bold.)

Model details

Technical Report

Training was performed using a fork of HRM-Text. Further details are provided in our technical report here

Memorisation audit

We run two audits independently from each other across four data categories: synthetic instruction-tuning data derived from agreement-backed sources (A), instruction-tuning data from Hugging Face sources with uncertain opt-out status (B), instruction-tuning data from Hugging Face sources with high confidence of no applicable opt-out (C), and other low-risk synthetic and reasoning post-training datasets (D). All remaining datasets are either covered by open licenses, in the public domain, or synthetic datasets derived from such openly licensed or public domain datasets. Our memorisation audit thus covers all data sources where memorisation risk is meaningful. Full details are provided in the appendix of the technical report. Targeted prefix attacks yielded predominantly short matches, with only 0.00021%–0.00073% of training documents have been retrieved with 50 tokens or more across categories.