Glossary
Temperature
Temperature is a setting that controls how predictable or adventurous an AI's wording is. Low temperature gives consistent, safe answers; high temperature gives varied, creative ones, plus a bit more nonsense.
Under the hood, an LLM doesn’t pick one “next word”; it weighs many candidates. Temperature decides how it chooses among them. Near zero: always take the safest bet, so the same question gets nearly the same answer every time. Higher: give unusual words a real chance, so answers get more varied and surprising.
Think of two cooks. Low temperature is the one who follows the recipe gram for gram, perfect when you need the invoice summary correct and repeatable. High temperature is the one who improvises: exactly who you want for “give me ten unusual names for my gelato flavors,” and exactly who you don’t want doing your taxes.
You can’t set temperature in the normal ChatGPT or Claude chat apps; the platforms pick sensible values for you. It’s a dial for builders using the API. It’s still worth knowing, because it explains why the same prompt gives different answers each time, and why “regenerate” is a legitimate strategy.
Where you’ll meet this
As a slider (typically 0 to 1 or 2) in API playgrounds from OpenAI, Anthropic, Google, and Mistral, in AI-app settings, and in developer documentation everywhere.