Researchers released Mimir v1, a 1-billion-parameter language model built on the Hierarchical Reasoning Model architecture and trained exclusively on ethically-sourced, permissibly-licensed post-training data drawn from a mixture of 161 datasets. Across 20 benchmarks spanning English, math, code, and Danish-language tasks, Mimir v1 outperforms the original HRM-Text 1B model and competes with larger models such as Qwen 3.5 4B, while also setting new results for Danish-language modeling. The model is publicly available on the Hugging Face Hub.
