Zurich

Higher values make the output more random, lower values make it more deterministic

Controls the cumulative probability threshold for nucleus sampling

Limits the number of tokens to consider for each generation step

Maximum number of tokens to generate in the response

Higher values stop the model from repeating the same info