Definition: Parameter that controls the randomness of LLM outputs. Lower values produce deterministic responses; higher values generate more creative, varied outputs.
— Source: NERVICO, Product Development Consultancy
What is Temperature
Temperature is a parameter that controls the randomness of responses generated by an LLM. Low values (0-0.3) produce more deterministic, predictable responses focused on the most likely option. High values (0.7-1.0) increase diversity and creativity in outputs, allowing the model to explore less probable options. It is one of the most important parameters when configuring an AI-based system.
How it works
When an LLM generates text, it computes a probability distribution over all possible tokens for the next position. Temperature modifies this distribution before selecting the token. With low temperature, probabilities concentrate on the most likely tokens, making selection nearly deterministic. With high temperature, probabilities are distributed more uniformly, giving less frequent tokens a better chance. A temperature of 0 always selects the most probable token, while a temperature of 1.0 maintains the model’s original distribution.
Why it matters
Temperature configuration has a direct impact on the quality and usefulness of an AI system’s responses. For tasks requiring precision and consistency (code generation, data extraction, factual responses), low temperature is essential. For creative tasks (content generation, brainstorming, idea exploration), higher temperature produces better results. Choosing the wrong temperature can cause hallucinations in contexts where precision is needed, or repetitive responses where variety is desired.
Practical example
A company configures two AI agents with different temperatures. The technical support agent operates at temperature 0.1 to guarantee consistent, precise responses based on documentation. The creative marketing agent operates at temperature 0.8 to generate variations of advertising copy. The result is that each agent is optimized for its specific use case, maximizing response quality in each context.
Related terms
- LLM - Language models where temperature is applied
- Hallucination - Risk that increases with high temperatures
- Grounding - Technique to mitigate excessive randomness
Last updated: February 2026 Category: Artificial Intelligence Related to: LLM, Sampling, AI Configuration, Hallucination Keywords: temperature, llm parameters, sampling, ai configuration, randomness, deterministic output, creative generation