Research

A British 7B model, properly documented.

We publish architecture, training data provenance, evaluation methodology, and safety work in full. The first technical paper on the Woven 7B model is in preparation.

We're working on a sovereign 7-billion-parameter model trained in Britain on UK compute. The technical paper covering pre-training data provenance, architecture decisions, evaluation methodology, and safety-tuning will be published here.

In the meantime, we publish operational details: the data centres our model runs in, the jurisdictional perimeter we maintain, and the audit standards we comply with.

First paper expected Q3 2026.