What Is Temperature?
A parameter controlling randomness in LLM output. Lower temperature (0.0-0.3) gives more deterministic responses, higher (0.7-1.0) gives more creative ones.
Temperature controls how 'creative' or 'random' a model's output is. At temperature 0, the model always picks the most likely next token, producing deterministic and predictable output. At temperature 1.0, it samples more broadly, producing varied and sometimes surprising responses.
For structured tasks like JSON generation, classification, or data extraction, use low temperature (0.1-0.3). For creative writing, brainstorming, or conversational AI, use higher temperature (0.7-0.9). Most benchmarks test at both ends to see how models perform across the spectrum.
Not all models support temperature. Reasoning models like GPT-5 and the o-series reject the temperature parameter entirely. Check our model profiles for quirks like this before building your integration.