Renaissance Labs

Atlas / Europe

Community collective · Aarhus, Denmark

Danish Foundation Models

A consortium of Danish universities building open language models, datasets and evaluations so that Danish is not left behind in AI.

language modelsDanishdatasetsevaluation

Aarhus, Denmark

Type
Community collective
Based in
Aarhus, Denmark
Key papers
3
Featured projects
8

About

Danish Foundation Models is a collaborative effort across Danish universities and research institutions to develop, evaluate and adapt open language AI for Danish society. The core partners are the University of Southern Denmark, Aarhus University through its Center for Humanities Computing, the University of Copenhagen and the Alexandra Institute.

Its members include professors Bolette Sandford Pedersen at Copenhagen, Kristoffer Nielbo at Aarhus and Peter Schneider-Kamp at Southern Denmark, with Kenneth Enevoldsen and many postdoctoral and PhD researchers. The project argues that smaller languages risk exclusion as language models become infrastructure, and it publishes models and datasets openly on Hugging Face.

The 2023 paper Danish Foundation Models set out the program. Danish Dynaword, released in 2025, is an openly licensed, continuously updated collection of Danish text derived from the public parts of the Danish Gigaword, with a companion post-training dataset, dfm-dyna-instruct.

Mission and ambitions

The consortium aims to give Danish society open, well-evaluated language technology that it can adopt and adapt. It treats openness of models and data as a condition for public benefit.

Known for

  1. 01

    Open Danish language models

  2. 02

    Danish Dynaword open text collection

  3. 03

    Scandinavian Embedding Benchmark

  4. 04

    Cross-university collaboration in Denmark

Milestones

  1. 2023

    Danish Foundation Models paper

    The project's founding paper laid out open foundation models for Danish.

  2. 2024

    Scandinavian Embedding Benchmarks

    Aarhus-led benchmark for text embeddings across Scandinavian languages.

  3. 2025

    Danish Dynaword

    Openly licensed Danish text collection that is updated continuously.

  4. 2025

    Dynaword paper

    Preprint describing continuously developed open datasets.

Open source

GitHubgithub.com/centre-for-humanities-computing
  • DaCy: The State of the Art Danish NLP pipeline using SpaCy (105 stars)

  • Tools for interactive visual exploration of semantic embeddings. (51 stars)

  • Blazing fast topic modelling for short texts. (37 stars)

  • A general-purpose NLP pipeline for Ancient Greek (29 stars)

Hugging Facehuggingface.co/danish-foundation-models

Research papers

Notable releases

Openness

Datasets, models and code are public; the Dynaword corpus is built from openly licensed sources.

People

Listed from the lab's own pages and public profiles.

Contact and community

Contact via the project website.

Sources

Compiled from the lab's public pages, Hugging Face and GitHub. Last checked 2026-11-17. Something wrong or out of date? Tell us on Discord.

Labs nearby