Temperature is a value that determines consistency in the model's output.
A high temperature will lead to creative answers, rather than deterministic behaviour.
Depending on the task, a low or a high temperature might be beneficial.