What Is Temperature in AI?

Temperature is a setting that controls how random or predictable a language model's output is. A lower temperature makes the model pick the most likely next words, giving focused, consistent responses. A higher temperature lets it choose less likely options more often, giving more varied and creative output.

How temperature works

When a model generates text, it assigns probabilities to possible next tokens. Temperature adjusts how sharply it favors the most probable one. At a low temperature, the model almost always takes the top choice, so its output is steady and repeatable. At a high temperature, the probabilities are flattened, so less likely tokens get picked more often, producing more surprising and diverse text.

At a temperature of zero, output is close to deterministic: the same prompt tends to give the same answer. As temperature rises, the same prompt yields more variation each time. There is no universally correct value; the right setting depends on whether the task wants reliability or variety.

Why temperature matters for AI

Temperature matters because it lets you tune output to the task without changing the model or the prompt. Work that needs accuracy and consistency, such as extracting data, following a format, or answering factual questions, calls for a low temperature, where the model stays on its most confident path. Creative work, such as brainstorming or drafting varied copy, benefits from a higher temperature that encourages novelty. Setting it too high on a precise task invites errors and drift, while setting it too low on a creative one makes output repetitive.