What language models does Burna AI use?

Burna AI runs on large language models from more than one established commercial provider, and a written routing policy assigns each task in the pipeline to a model. The properties that matter for a trial record, output constrained to valid CTCAE terms and grades, mandatory citation, per-drug WHO-UMC and Kramer attribution and clinician sign-off, belong to the pipeline rather than to any one model, so a model can be upgraded without changing what a grade must carry. Contracts with model providers exclude customer data from training, and the current list of subprocessors is at burna.ai/subprocessors.