Temperature (AI Temperature)

Temperature is a parameter that controls the randomness of a language model's output by scaling the probability distribution over possible next tokens before sampling. A temperature of 0 makes the model deterministic (always choosing the highest-probability token), while higher values (e.g., 0.7–1.0) increase diversity and creativity by flattening the distribution and making lower-probability tokens more likely to be selected. At very high temperatures, output becomes largely random and incoherent.

Temperature is a practical tuning knob in nearly every production LLM application and must be set thoughtfully for each use case. For tasks requiring factual accuracy and consistency — such as structured data extraction, classification, or policy-compliant customer communications — low temperatures (0.0–0.3) reduce variability and hallucination risk. For creative tasks like marketing copy generation, product naming, or brainstorming — where diversity is valuable — moderate to higher temperatures (0.7–1.0) produce more varied and interesting outputs. Most commerce AI systems expose temperature as a configurable parameter, and getting it right for each application is a key part of prompt and system design.