Temperature
Why does AI answer the same question differently every time?
Why does AI answer the same question differently every time?
All 3 runs are identical: highly predictable, ideal for code, JSON, and structured extraction.
Low: steady and predictable: Asking the same question repeatedly tends to converge; suited to fixed formats, code generation, and extraction where stability matters.
High: varied and divergent: The same question gets different phrasings each time; suited to brainstorming, naming, and exploring options.
It changes randomness, not ability: Lowering temperature does not make the model smarter, and raising it does not either; it only affects which candidate is sampled.
This task needs stable output: set the temperature to its lowest setting and return JSON in the fields and format I specify. Run the same input three times after each change and confirm the structure is identical before delivering.