# Temperature

> A sampling setting that controls how random a language model's output is — low values pick the most likely words, high values take more chances.

A model does not emit one answer, it emits a probability for every possible next token. Temperature reshapes those probabilities before one is sampled: near zero the model almost always takes the front-runner, near one it explores. Same model, same prompt, different personality.

The setting should follow the task. Support answers, extraction and anything a user will fact-check want low temperature; brainstorms and first drafts can take more. What temperature never does is create knowledge — a high setting makes a wrong answer more creative, not more right.

## Related terms

- https://dfieldsolutions.com/en/glossary/llm.md
- https://dfieldsolutions.com/en/glossary/hallucination.md
- https://dfieldsolutions.com/en/glossary/system-prompt.md

---

Source: https://dfieldsolutions.com/en/glossary/temperature
DField Solutions — Dunakeszi, Hungary — dezso@dfieldsolutions.com
Booking: see https://dfieldsolutions.com/en/contact