Glossary term
Glossary term
Architecture
Parameter controlling randomness in model output generation.
A hyperparameter that controls the degree of randomness of a model's output. Higher temperatures result in more random output, while lower temperatures result in less random output.
Choosing the best temperature depends on the specific application and or string values.
GitHub Copilot uses temperature=0.2 for code completion to favour deterministic, correct outputs - higher temperature would introduce creative but potentially incorrect code variations that frustrate developers.
Character.AI uses temperature=1.0+ for creative roleplay applications - higher temperature produces more varied, surprising responses that feel less robotic, improving user engagement metrics.
A contract-review application uses temperature=0 (greedy decoding) to ensure identical outputs for identical inputs - enabling reproducible, auditable AI decisions required by the company's ISO 42001 governance framework.
Created for this library
A creative writing tool uses higher temperature for brainstorming and lower temperature for editing tasks.
A code-completion vendor uses a low temperature for code suggestions so outputs are deterministic and reproducible.
An LLM product team standardizes temperature defaults per feature to keep generation behavior consistent across product surfaces.
Definition source: Google for Developers Machine Learning Glossary | Creative Commons Attribution 4.0 License