In generative AI, temperature is a parameter that controls the randomness of output. Low temperatures yield more deterministic and focused results, while high temperatures produce more diverse and creative — but potentially less accurate — responses.