Several parameters control the randomness, length, and style of API responses.

Key Parameters

• temperature (0–2): Controls randomness. 0 = deterministic, 0.7 = balanced, 1.5+ = very creative
• max_tokens: Maximum tokens in the response. Controls cost and length.
• top_p: Alternative to temperature — nucleus sampling (usually leave at 1)
• frequency_penalty (0–2): Reduces repetition of the same words
• presence_penalty (0–2): Encourages the model to talk about new topics

Common Settings

• Factual/code tasks: temperature 0–0.2
• Balanced: temperature 0.7
• Creative writing: temperature 1.0–1.2

Reference:

API parameters reference

https://en.wikipedia.org/wiki/Special:Search?search=Temperature%20and%20Parameters