UK-LLM, the U.K. sovereign AI initiative, is building a new AI model for Welsh based on NVIDIA Nemotron. The model is designed to reason in both English and Welsh and could support public services in healthcare, education and legal resources.
Welsh is spoken by about 850,000 people in Wales today. The project is led by University College London, with Bangor University and NVIDIA working on the model. U.K. Prime Minister Keir Starmer said the effort will help make public services accessible in Welsh and protect cultural heritage.
The UK-LLM project was established in 2023 as BritLLM and has previously released two models for U.K. languages. Its new Welsh model aligns with Welsh government efforts under Cymraeg 2050, which aims to reach a million speakers by 2050.
Nscale will make the model available to developers through its application programming interface. The model, along with the Welsh training and evaluation datasets, is expected to be made available for enterprise and public sector use.
To build the system, the team used the 49-billion-parameter Llama Nemotron Super model and 9-billion-parameter Nemotron Nano model, post-training them on Welsh-language data. Because Welsh has less available source data than languages such as English or Spanish, the team used NVIDIA NIM microservices for gpt-oss-120b and DeepSeek-R1 to translate NVIDIA Nemotron open datasets with over 30 million entries from English to Welsh.
The team used a GPU cluster through the NVIDIA DGX Cloud Lepton platform and hundreds of NVIDIA GH200 Grace Hopper Superchips on Isambard-AI, the U.K.’s most powerful supercomputer, based at University of Bristol and backed by £225 million in government investment.
Gruffudd Prys, senior terminologist and head of the Language Technologies Unit at Canolfan Bedwyr, said the aim is to keep Welsh “a living, breathing language that continues to develop with the times.” He said AI has potential to help with second-language acquisition and to improve native speakers’ language skills.
Pontus Stenetorp, professor of natural language processing and deputy director for the Centre of Artificial Intelligence at University College London, said the collaboration enabled the team to create new training data and train a new model in record time.
The model is intended to help public institutions and businesses operating in Wales translate content or provide bilingual chatbot services. The UK-LLM team also plans to apply the same approach to other U.K. languages, including Cornish, Irish, Scots and Scottish Gaelic.
Source: blogs.nvidia.com.
Companies can share verified announcements through Newz9’s international press release submission page.
