Term
Temperature
A sampling parameter that scales the model's output probabilities before sampling. Lower values (0.0–0.3) produce focused, repeatable outputs; higher values (0.7–1.2) produce more varied, creative outputs. Most coding tools default to 0.0–0.3.
Background
Temperature mathematically divides the logits by T before softmax. At T=0 the model always picks the most-likely next token (greedy). At T=1 the distribution is sampled as the model originally calibrated it. Higher T flattens the distribution and increases randomness. For code generation, low temperatures dominate because deterministic, syntactically correct outputs are what users want. For brainstorming or natural-language generation, higher T is reasonable.