The first release of the Munin models by the Danish Foundation Models project, being existing base models post-trained for Danish and English.
AI & ML interests
None defined yet.
Recent Activity
View all activity
Benchmarks used and maintained by Danish Foundation Models.
-
EuroEval Leaderboard
📊13The robust European language model benchmark.
-
ScandEval: A Benchmark for Scandinavian Natural Language Processing
Paper • 2304.00906 • Published • 4 -
MTEB Leaderboard
📊7.7kEmbedding Leaderboard
-
The Scandinavian Embedding Benchmarks: Comprehensive Assessment of Multilingual and Monolingual Text Embedding
Paper • 2406.02396 • Published • 1
A collection of EuroEval compatible datasets which can be run using: `euroeval --dataset {dataset name} --model {model name}`
Papers related to Danish Foundation Models
-
DFM Mimir v1: An Open HRM Delivering Frontier Performance at 1B Parameters Using Only Permissible Post-Training Data
Paper • 2608.13517 • Published • 33 -
Dynaword: From One-shot to Continuously Developed Datasets
Paper • 2508.02271 • Published • 15 -
Last Translation Benchmark
Paper • 2609.04173 • Published • 43 -
MVEB: Massive Video Embedding Benchmark
Paper • 2606.14958 • Published • 16
These models constitute state-of-the-art models for Danish within their respective domain (highlighted below the model).
-
mistralai/Mistral-Small-3.1-24B-Instruct-2503
24B • Updated • 509k • 1.38k -
google/gemma-3-27b-it
Image-Text-to-Text • 27B • Updated • 411k • • 2.04k -
google/gemma-3n-E4B-it
Image-Text-to-Text • 8B • Updated • 15.5k • • 928 -
google/gemma-2-9b-it
Text Generation • 9B • Updated • 1.14M • • 1.02k
A collection of dynawords, target various languages
Datasets related to AI-Arenaen
-
danish-foundation-models/ai-arenaen-conversations
Viewer • Updated • 3.09k • 18 • 1 -
danish-foundation-models/ai-arenaen-reactions
Preview • Updated • 1 • 1 -
danish-foundation-models/ai-arenaen-votes
Viewer • Updated • 1.77k • 1 • 1 -
ministere-culture/comparia-conversations
Viewer • Updated • 410k • 185 • 72
This is a collection of artifact released as a part of the paper: "Dynaword: From One-shot to Continuously Developed Datasets".
-
Dynaword: From One-shot to Continuously Developed Datasets
Paper • 2508.02271 • Published • 15 -
danish-foundation-models/danish-dynaword
Viewer • Updated • 22.2M • 11.9k • 23 -
danish-foundation-models/gemma-3-1b-cpt-dynaword-matched-v1
Text Generation • 1.0B • Updated • 15 -
danish-foundation-models/gemma-3-1b-scratch-dynaword-full-v1
Text Generation • 1.0B • Updated • 19
These include high-quality Danish text datasets for pre-training, fine-tuning, etc.
The first release of the Munin models by the Danish Foundation Models project, being existing base models post-trained for Danish and English.
A collection of dynawords, target various languages
Benchmarks used and maintained by Danish Foundation Models.
-
EuroEval Leaderboard
📊13The robust European language model benchmark.
-
ScandEval: A Benchmark for Scandinavian Natural Language Processing
Paper • 2304.00906 • Published • 4 -
MTEB Leaderboard
📊7.7kEmbedding Leaderboard
-
The Scandinavian Embedding Benchmarks: Comprehensive Assessment of Multilingual and Monolingual Text Embedding
Paper • 2406.02396 • Published • 1
Datasets related to AI-Arenaen
-
danish-foundation-models/ai-arenaen-conversations
Viewer • Updated • 3.09k • 18 • 1 -
danish-foundation-models/ai-arenaen-reactions
Preview • Updated • 1 • 1 -
danish-foundation-models/ai-arenaen-votes
Viewer • Updated • 1.77k • 1 • 1 -
ministere-culture/comparia-conversations
Viewer • Updated • 410k • 185 • 72
A collection of EuroEval compatible datasets which can be run using: `euroeval --dataset {dataset name} --model {model name}`
This is a collection of artifact released as a part of the paper: "Dynaword: From One-shot to Continuously Developed Datasets".
-
Dynaword: From One-shot to Continuously Developed Datasets
Paper • 2508.02271 • Published • 15 -
danish-foundation-models/danish-dynaword
Viewer • Updated • 22.2M • 11.9k • 23 -
danish-foundation-models/gemma-3-1b-cpt-dynaword-matched-v1
Text Generation • 1.0B • Updated • 15 -
danish-foundation-models/gemma-3-1b-scratch-dynaword-full-v1
Text Generation • 1.0B • Updated • 19
Papers related to Danish Foundation Models
-
DFM Mimir v1: An Open HRM Delivering Frontier Performance at 1B Parameters Using Only Permissible Post-Training Data
Paper • 2608.13517 • Published • 33 -
Dynaword: From One-shot to Continuously Developed Datasets
Paper • 2508.02271 • Published • 15 -
Last Translation Benchmark
Paper • 2609.04173 • Published • 43 -
MVEB: Massive Video Embedding Benchmark
Paper • 2606.14958 • Published • 16
These include high-quality Danish text datasets for pre-training, fine-tuning, etc.
These models constitute state-of-the-art models for Danish within their respective domain (highlighted below the model).
-
mistralai/Mistral-Small-3.1-24B-Instruct-2503
24B • Updated • 509k • 1.38k -
google/gemma-3-27b-it
Image-Text-to-Text • 27B • Updated • 411k • • 2.04k -
google/gemma-3n-E4B-it
Image-Text-to-Text • 8B • Updated • 15.5k • • 928 -
google/gemma-2-9b-it
Text Generation • 9B • Updated • 1.14M • • 1.02k