Learning Objectives:
(1) Define Temperature and its operational range.
(2) Predict output behavior at low, medium, and high Temperature values.
(3) Select the correct Temperature for 5 different use cases.
What Is Temperature?
Temperature controls the randomness and diversity of the model’s output. Mathematically, it scales the probability distribution of next-token selection — making high-probability tokens even more dominant (low temperature) or flattening the distribution so lower-probability tokens get more chances (high temperature).
MENTAL MODEL | Temperature is like the creativity dial on a writer. Temperature = 0 is a precise technical writer who always says the safest, most expected thing. Temperature = 1 is a creative director who surprises you with unexpected word choices and imaginative ideas. |
Temperature × Use Case Decision Matrix
Temperature range and behavior
Use Case | Recommended Temperature | Why |
Code Generation | 0.2 | Correctness over creativity — syntax must be exact |
Data Analysis Scripts | 0.2 | Deterministic, follows established patterns |
Code Comments | 0.3 | Slightly varied but still precise and relevant |
Chatbot Responses | 0.5 | Natural conversation — coherent but not robotic |
Marketing / Copywriting | 0.7 | Diverse vocabulary, punchy and engaging |
Creative Writing / Stories | 0.7–0.8 | Exploratory, imaginative, emotionally resonant |
Brainstorming / Ideation | 0.9–1.0 | Maximum diversity of ideas — quantity over precision |