As AI adoption expands globally, people across languages and cultures are using it in more aspects of their daily lives. General-purpose language models serve as a powerful foundation, but they sometimes lack the local knowledge, cultural nuance, and precision a particular application needs.

Post-training can adapt an open model to these needs. With suitable data and evaluation, it can improve accuracy and usefulness while supporting safer interactions with the model. This gives countries and industries greater control over how their agents reflect local priorities and serve the people who use them.

NVIDIA built Nemotron-Personas-France with French AI startup Pleias, and Nemotron-Personas-Belgium with Pleias and Belgian university KU Leuven.…