Open LLM Research from the Netherlands, for the World
We push the frontier of large language model research in the open — sharing everything we build, learn, and discover.
Why We Exist
The frontier of AI is increasingly concentrated in a handful of large labs. Critical research happens behind closed doors, models ship without published weights, and the broader community is left to catch up. We think that's the wrong direction.
At openoranje, we conduct and publish LLM research entirely in the open — models, datasets, training code, and findings. We are based in the Netherlands but our scope is global: any language, any domain, any architecture worth exploring.
The Dutch tradition of openness, pragmatism, and building things that work guides how we operate. Science should be reproducible. AI should be a public good.
Our Values
These principles guide everything we do at openoranje.
Reproducible Science
Every result we publish comes with the code, data, and methodology to reproduce it. No black boxes.
Open Source First
All our models, datasets, and training code are freely available. We believe in building together.
No Boundaries on Scope
Dutch, English, multilingual, multimodal — we follow the research wherever it leads, without artificial constraints.
Accessible AI
Research and models should be usable by anyone — not gated behind expensive APIs or proprietary licenses.
Community Driven
We welcome collaborators, contributors, and critics. The best research happens in the open.
Research Excellence
We hold ourselves to a high bar — rigorous evaluation, honest reporting of failures, and peer-reviewable methodology.