Temperature controls the model's responses: a low setting yields more predictable and consistent results, while a high setting allows for greater creativity and variability in the outputs.

Max tokens sets the limit on the number of words or characters the model can generate in a single response, controlling the length of the output.

Top_p limits the model to choose from the most likely words, making sure the responses are varied yet sensible.