Temperature: Temperature is a setting that controls how random a language model's output is. Low values make the model pick the most likely next word and give consistent, predictable answers; higher values allow more varied and creative responses.
When a model writes, it holds a list of possible next tokens ranked by probability. Temperature decides how strictly it follows that ranking. At a temperature near zero it almost always takes the top choice, so the same prompt gives nearly the same answer every time. Turn it up and the model samples from lower-ranked options too, producing more surprise and more risk of nonsense.
Most chat apps do not expose temperature; the vendor picks a moderate default. You will meet it when you use an API, a developer playground, or an automation platform. Typical scales run from 0 to 1, and some providers allow up to 2. Check the vendor's documentation for the exact range.
Use low temperature for anything that must be repeatable: data extraction, classification, code, and formatted output that feeds another system. Use a higher setting for brainstorming names, headlines, or alternative angles, where you want ten different ideas instead of the same safe one ten times. Temperature does not make a model more accurate; it only changes how adventurous it is.
Example at work
A support team's automation classifies incoming tickets by category and urgency. The engineer sets temperature to 0 so identical tickets always get identical labels. For the weekly newsletter, the same team runs a separate prompt at a higher temperature to generate a dozen subject-line options to pick from.
Why it matters
If you ever build or buy an automated workflow on top of an AI model, temperature is one of the first settings to ask about. It explains why a model gives different answers to the same question and how to make it stop doing that when consistency matters.
Related terms
- InferenceInference is the act of running a trained AI model to produce an output, such as answering a prompt or classifying an image. Training happens once; inference happens every time anyone uses the model, and it is what you pay for as a user.
- API (application programming interface)An API (application programming interface) is a way for one piece of software to talk to another. An AI API lets your own tools, spreadsheets, or automations send a prompt to a model and receive the response, without a person typing in a chat window.
- Large language model (LLM)A large language model (LLM) is a neural network trained on enormous amounts of text to predict the next word, then tuned to follow instructions and hold a conversation. LLMs power ChatGPT, Claude, Gemini, and Microsoft Copilot.
- TokenA token is the unit of text a language model reads and writes, usually a word or a piece of a word. Models measure input limits, output limits, and pricing in tokens; as a rough rule, 1,000 tokens is about 750 words of English.
- Workflow automationWorkflow automation connects apps and steps so that a trigger, such as a new form response or an incoming email, automatically runs a sequence of actions, increasingly with an AI step that reads, decides, or writes along the way.