APREX · EST. 2026Norwegian and Nordic Language Models
A complete, regularly updated overview of large language models (LLMs) trained on Norwegian, Swedish, Danish, Finnish, and Icelandic — covering open-weights releases, licences, and recommended use cases.
2026-05-25 10:00 CET
| Model | Country | Languages | Open weights | License | Best for |
|---|---|---|---|---|---|
NorBERT 3 by Universitetet i Oslo (UiO) | Norway | no, nn, en | Yes | MIT | Klassifisering, Navngitt enhetsgjenkjenning (NER), Søk- og relevansanalyse |
NbGPT by Nasjonalbiblioteket | Norway | no, nn, en, se | Yes | Apache-2.0 | Tekstgenerering, Oppsummering på norsk, Kulturell kontekstforståelse |
GPT-SW3 by AI Sweden | Sweden | sv, no, da, is, en | Yes | AI Sweden License | Svensk og skandinavisk tekstbehandling, Spesialisert innholdsgenerering, Offentlig sektor i Sverige |
Viking-7B by Silo AI / SiloGen | Finland | fi, sv, no, da, is, en | Yes | Apache-2.0 | Nordisk flerspråklig oversettelse, Generell tekstskriving på finsk og svensk |
Poro 34B by Silo AI / TurkuNLP | Finland | fi, en | Yes | Apache-2.0 | Finsk og engelsk tekstgenerering, Forskning på lavressursspråk, Lokal kjøring av åpne modeller |
Sources & methodology.
Entries are structured by 24AI and validated against official or primary sources where provided. Strategic datasets may be manually reviewed by Aprex when clearly marked. Each entry carries source context and a confidence score.
- NorBERT 3 GitHub Repository· Language Technology Group (UiO)research
- Nasjonalbiblioteket AI-satsing· Nasjonalbiblioteketofficial
- GPT-SW3 Model Release Info· AI Swedenofficial
- Silo AI Viking Model Launch· Silo AIcompany
- Poro-34B Model Card (Hugging Face)· SiloGen / TurkuNLP / HPLTresearch
Frequently Asked Questions.
Why do Norway and the Nordic countries need their own language models?
Global models such as ChatGPT are trained predominantly on English-language material and often lack deep understanding of Norwegian culture, legislation, social structures, and local nuances. Norwegian models trained on collections from the National Library of Norway (Nasjonalbiblioteket) and similar institutions help preserve Nordic languages in the AI era and enable organisations to run AI locally on their own infrastructure.
Are Norwegian and Nordic language models free to use?
Most models released by the University of Oslo (UiO), the National Library of Norway, AI Sweden, and Silo AI are published under open licences such as Apache 2.0 or MIT and can be downloaded for free from Hugging Face. A small number of models are available only via a commercial API.
Can international researchers and media cite this dataset?
Yes. The Norwegian and Nordic Language Models dataset is published by 24AI under Creative Commons Attribution (CC BY 4.0). You are free to share and adapt the data as long as proper attribution is given to 24AI / Aprex AS.