AI & LLMs
Temperature
The setting that controls how predictable vs. varied an LLM's next-word choices are.
Reviewed by the RadarTrek editorial team · June 2026
Temperature reshapes the probability distribution the model samples from when choosing its next token. At temperature 0, the model almost always picks the single most likely token — useful for extraction and classification where you want the same answer every time. Higher temperatures let less-likely tokens compete, producing more varied, creative output at the cost of consistency.
Why it matters
- —Temperature 0 is the right default for data extraction, classification, and anything needing repeatable output.
- —Higher temperatures (0.7-1.0) suit creative writing and brainstorming where variation is the point, not a bug.
- —Temperature 0 is "almost" deterministic, not perfectly — floating-point differences can still cause tiny variation.
Where to learn this
🎓
Temperature and Sampling
How LLMs Actually Work course
This is the exact lesson that covers this term in depth — with examples, diagrams, and a hands-on exercise.