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.</p>\n","updatedAt":"2026-08-17T12:18:15.454Z","author":{"_id":"6652354cb88e4539b2189cd7","avatarUrl":"https://cdn-avatars.huggingface.co/v1/production/uploads/6652354cb88e4539b2189cd7/kZ7Mi6Yz7zbOSLqgFW5jt.jpeg","fullname":"Gianluca Barmina","name":"giannor","type":"user","isPro":false,"isHf":false,"isHfAdmin":false,"isMod":false,"followerCount":6,"isUserFollowing":false}},"numEdits":0,"identifiedLanguage":{"language":"en","probability":0.8945193290710449},"editors":["giannor"],"editorAvatarUrls":["https://cdn-avatars.huggingface.co/v1/production/uploads/6652354cb88e4539b2189cd7/kZ7Mi6Yz7zbOSLqgFW5jt.jpeg"],"reactions":[],"isReport":false}}],"primaryEmailConfirmed":false,"paper":{"id":"2608.13517","authors":[{"_id":"6a7ec24b42823931a1f177db","name":"Peter Schneider-Kamp","hidden":false},{"_id":"6a7ec24b42823931a1f177dc","name":"Jacob Nielsen","hidden":false},{"_id":"6a7ec24b42823931a1f177dd","user":{"_id":"6652354cb88e4539b2189cd7","avatarUrl":"https://cdn-avatars.huggingface.co/v1/production/uploads/6652354cb88e4539b2189cd7/kZ7Mi6Yz7zbOSLqgFW5jt.jpeg","isPro":false,"fullname":"Gianluca Barmina","user":"giannor","type":"user","name":"giannor"},"name":"Gianluca Barmina","status":"claimed_verified","statusLastChangedAt":"2026-08-15T16:45:05.206Z","hidden":false},{"_id":"6a7ec24b42823931a1f177de","name":"Kenneth Enevoldsen","hidden":false},{"_id":"6a7ec24b42823931a1f177df","name":"Lukas Galke Poech","hidden":false}],"publishedAt":"2026-08-13T00:00:00.000Z","submittedOnDailyAt":"2026-08-17T00:00:00.000Z","title":"DFM Mimir v1: An Open HRM Delivering Frontier Performance at 1B Parameters Using Only Permissible Post-Training Data","submittedOnDailyBy":{"_id":"6652354cb88e4539b2189cd7","avatarUrl":"https://cdn-avatars.huggingface.co/v1/production/uploads/6652354cb88e4539b2189cd7/kZ7Mi6Yz7zbOSLqgFW5jt.jpeg","isPro":false,"fullname":"Gianluca Barmina","user":"giannor","type":"user","name":"giannor"},"summary":"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 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. The model is available on the Hugging Face Hub: https://huggingface.co/danish-foundation-models/DFM-Mimir","upvotes":16,"discussionId":"6a7ec24b42823931a1f177e0","projectPage":"https://www.foundationmodels.dk","githubRepo":"https://github.com/schneiderkamplab/HRM-Text","githubRepoAddedBy":"user","ai_summary":"Mimir v1 is a 1-billion-parameter Hierarchical Reasoning Model trained solely on permissible data that achieves competitive English results and state-of-the-art Danish performance across multiple benchmarks.","ai_keywords":["Hierarchical Reasoning Model","HRM architecture","language model","post-training data","benchmarks"],"ai_summary_model":"thinkingmachines/Inkling-Small","githubStars":0,"organization":{"_id":"68adbfb1dd070a92488069b1","name":"SDU-Denmark","fullname":"University of Southern Denmark (SDU)","avatar":"https://cdn-avatars.huggingface.co/v1/production/uploads/68adbdb50fdaa186aa43d1ce/f1kfMH47RIckIAOWEi1mv.png"}},"canReadDatabase":false,"canManagePapers":false,"canSubmit":false,"hasHfLevelAccess":false,"upvoted":false,"upvoters":[{"_id":"68b031d6aa3a9d6ef8ff91ca","avatarUrl":"https://cdn-avatars.huggingface.co/v1/production/uploads/no-auth/-uFUU2OfVN02ttCtgIVOw.png","isPro":false,"fullname":"Annemette Brok Pirchert","user":"popunicorn","type":"user"},{"_id":"6652354cb88e4539b2189cd7","avatarUrl":"https://cdn-avatars.huggingface.co/v1/production/uploads/6652354cb88e4539b2189cd7/kZ7Mi6Yz7zbOSLqgFW5jt.jpeg","isPro":false,"fullname":"Gianluca Barmina","user":"giannor","type":"user"},{"_id":"62cd65da816d30201adca921","avatarUrl":"https://cdn-avatars.huggingface.co/v1/production/uploads/62cd65da816d30201adca921/M5b4wXhYokQjLzEoGcBmI.jpeg","isPro":false,"fullname":"Lukas Galke Poech","user":"lgalke","type":"user"},{"_id":"65857e9d3af0da4f5ebb256d","avatarUrl":"https://cdn-avatars.huggingface.co/v1/production/uploads/65857e9d3af0da4f5ebb256d/CMPg14q1D8yDf0eZxZ25E.png","isPro":false,"fullname":"Danial Namazifard","user":"namazifard","type":"user"},{"_id":"65e44b1927dc8aa4705b4f64","avatarUrl":"https://cdn-avatars.huggingface.co/v1/production/uploads/65e44b1927dc8aa4705b4f64/ibZNEhBLjuJR89BhU52XI.png","isPro":false,"fullname":"William Brach","user":"williambrach","type":"user"},{"_id":"60d33fbbd7b174177faabd4f","avatarUrl":"https://cdn-avatars.huggingface.co/v1/production/uploads/60d33fbbd7b174177faabd4f/pfyv_xj2B2m2N4F4sT9zJ.jpeg","isPro":true,"fullname":"Mike Zhang","user":"jjzha","type":"user"},{"_id":"65dee4eb2df2dd7ceecb5850","avatarUrl":"https://cdn-avatars.huggingface.co/v1/production/uploads/65dee4eb2df2dd7ceecb5850/WZCx-1X-7944O-BX7h29L.jpeg","isPro":false,"fullname":"Jacob Nielsen","user":"JacobBITLABS","type":"user"},{"_id":"6a82fe44908e3056cfa04fe9","avatarUrl":"/avatars/f62b588292ce9f06569c2811cd195326.svg","isPro":false,"fullname":"Henrik Hoffmann","user":"henrikhoffmanndk","type":"user"},{"_id":"5ff5943752c26e9bc240bada","avatarUrl":"https://cdn-avatars.huggingface.co/v1/production/uploads/5ff5943752c26e9bc240bada/Exyzf3C_gJ2KdsL4K5_cq.png","isPro":false,"fullname":"Kenneth C. Enevoldsen","user":"KennethEnevoldsen","type":"user"},{"_id":"64d02f09a2e7f9ff61035013","avatarUrl":"https://cdn-avatars.huggingface.co/v1/production/uploads/64d02f09a2e7f9ff61035013/V9HK7GVnpBwenNz8SSyus.jpeg","isPro":false,"fullname":"Yevhen Kostiuk","user":"yevhenkost","type":"user"},{"_id":"6a544b59c84594e3ad97a7e3","avatarUrl":"https://cdn-avatars.huggingface.co/v1/production/uploads/6a544b59c84594e3ad97a7e3/or4F9W2-FVx1Kj1wgRsd7.jpeg","isPro":true,"fullname":"Magnus Lundstedt","user":"maglun","type":"user"},{"_id":"63ea5843c9d1e22a28c03914","avatarUrl":"/avatars/37f5aa271dfff6556a07fb273b20680a.svg","isPro":false,"fullname":"Peter Schneider-Kamp","user":"peter-sk","type":"user"}],"acceptLanguages":["en"],"dailyPaperRank":0,"organization":{"_id":"68adbfb1dd070a92488069b1","name":"SDU-Denmark","fullname":"University of Southern Denmark (SDU)","avatar":"https://cdn-avatars.huggingface.co/v1/production/uploads/68adbdb50fdaa186aa43d1ce/f1kfMH47RIckIAOWEi1mv.png"},"markdownContentUrl":"https://huggingface.co/buckets/huggingchat/papers-content/resolve/2608/2608.13517.md","query":{}}">
DFM Mimir v1: An Open HRM Delivering Frontier Performance at 1B Parameters Using Only Permissible Post-Training Data
Abstract
Mimir v1 is a 1-billion-parameter Hierarchical Reasoning Model trained solely on permissible data that achieves competitive English results and state-of-the-art Danish performance across multiple benchmarks.
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 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. The model is available on the Hugging Face Hub: https://huggingface.co/danish-foundation-models/DFM-Mimir
Community
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.
Upload images, audio, and videos by dragging in the text input, pasting, or clicking here.
Tap or paste here to upload images
Cite arxiv.org/abs/2608.13517 in a dataset README.md to link it from this page.
Cite arxiv.org/abs/2608.13517 in a Space README.md to link it from this page.
Discussion (0)
Sign in to join the discussion. Free account, 30 seconds — email code or GitHub.
Sign in →No comments yet. Sign in and be the first to say something.