Temperature

Temperature is a parameter that controls how random a language model's output is. At temperature 0, the model always picks the most probable next token—producing deterministic, repetitive, but reliable output. At temperature 1, the model samples more broadly from its probability distribution—producing varied, creative, but less predictable output. Higher temperatures (above 1) make the model increasingly random, often incoherently so. The right temperature depends entirely on the task. For extracting structured data from documents, you want temperature 0—you need the same answer every time. For brainstorming marketing copy variations, you want 0.7 to 0.9—you need diversity. Most API defaults sit around 0.7, which is a reasonable middle ground that few teams bother to adjust. Tuning temperature is one of those small decisions that meaningfully affects output quality and is trivially easy to get right once you understand what it does.