GPT-NL

GPT-NL is a Dutch national language model developed by the applied research organisation TNO, the Netherlands Forensic Institute, and SURF, funded with 13.5 million euros from the Dutch Ministry of Economic Affairs. The project aims for a transparent, fair, and verifiable model aligned with Dutch and European values, trained on the Snellius national supercomputer using exclusively lawfully sourced data, including a paid collective licensing agreement with Dutch news publishers.

Beta pilots began in late February 2026 across municipalities, civil-service drafting tools, and forensic data classification, with a broader rollout planned for later in the year. Early results report the model beating GPT-3 on Dutch summarisation benchmarks. Access is currently through the consortium's pilot program rather than public weights or a general API.