loading
loading
A setting that controls how random or deterministic a model's outputs are.
At temperature 0, the model consistently picks the most probable next token, making outputs highly deterministic and suitable for code or structured data. Higher temperatures (0.7-1.0) introduce variability, which benefits creative tasks but can make factual responses less reliable. Most production agentic systems run at low temperature for predictability.
Plainly
Think of Temperature as the brain part that guesses or decides. A setting that controls how random or deterministic a model's outputs are.
In practice
Use it when model choice, prompts, latency, cost, or quality affect the product result. In practice, define the owner, input, output, and failure mode before you rely on it.