Completions are model-generated outputs produced in response to a prompt or partial input. In language models, a completion typically continues, answers, or transforms the input based on patterns learned during training and any provided context.