Danish Foundation Models (Munin / DFM)
Open-source Danish and Scandinavian foundation models and encoders trained on authentic Danish corpora.
About
The Danish Foundation Models (DFM) project, initiated by the Center for Humanities Computing at Aarhus University in collaboration with national research partners, provides open-weights foundation models and embeddings specifically trained on high-quality Danish texts, parliamentary records, Danish news, and cultural heritage data. Notable releases include the Munin decoder family and Danish RoBERTa/T5 variants hosted on Hugging Face.
Actionable Use Cases
- 01Generate natural Danish text and structured responses using Hugging Face Transformers
- 02Compute high-dimensional semantic embeddings optimized for Danish RAG applications
- 03Perform sentiment analysis and grammatical parsing on historical and modern Danish
Technical Details & Limits
Weights are freely accessible on Hugging Face under Apache 2.0 / Open RAIL licenses. Can be run locally via Ollama, vLLM, or Transformers.
Build Access with Resources
Turn this API into an MCP server or CLI for Claude, Cursor, and terminal agents:
Instant MCP & CLI generator from OpenAPI, Swagger & APIs
High-level Python & TypeScript framework for building MCP servers
Turn entire websites and portals into clean, LLM-ready Markdown
Something wrong, outdated, or missing for Danish Foundation Models (Munin / DFM)?