GLOSSARY
Temperature
The dial that controls how predictable or creative a model's output is — low for factual precision, high for variety.
When a model generates text it assigns probabilities to possible next tokens, then samples. Temperature reshapes those probabilities: at 0 the model almost always picks the single most likely token (deterministic, repetitive, safe); at 1.0 and above it samples more freely (varied, creative, occasionally incoherent).
Practical rule: temperature near 0 for tasks with a right answer — extraction, classification, code patches, SQL. Higher values (0.7–1.0) for brainstorming, naming, fiction and anywhere surprise is the point. Most chat apps fix this for you; API users should set it deliberately.