Summarise this article with AI
If you send your requests to tools such as ChatGPT, Claude or Gemini through an API, one of the values you supply is the “temperature” parameter. It controls how creative and how variable the model’s answers will be. The parameter determines how individual tokens (words or parts of words) are picked while the text is being generated. Temperature can take a value between 0 and 1, where:
- A low value (close to 0): The model produces conservative, more predictable answers. Token selection is more deterministic, meaning the model favours the most probable options. A value of 0.1, for example, will make the model write very consistent and rather uncreative text.
- A high value (close to 1): The model produces more creative and more varied answers. Token selection is more random, meaning the model will more often reach for less probable options. A value of 0.9, for example, will make the model write more variable and less predictable text.
Examples of use:
- Value 0.7: This middle setting is used very often, because it offers a good compromise between creative and consistent answers. The model is creative enough to produce interesting text while keeping a certain level of consistency and relevance.
- Value 1.0: For when you want maximum creativity and accept a higher risk of answers that make less sense. In this mode the model experiments more and may produce original but occasionally incoherent or nonsensical text.
- Value 0.0: When you need the most accurate and most predictable answer possible, use a value close to 0. The model will stick strictly to the most probable options, which leads to less creative but very consistent results.
