Higher values make the response more random, lower values make it more focused and deterministic.

The model considers the tokens with TopP probability mass. So 0.1 means only the tokens comprising the top 10% probability mass are considered.

The maximum number of tokens that the model can use in it's response.