Temperature
A dial for how adventurous the model is — low is focused and repetitive, high is creative and risky.
Think of it like
The difference between a careful accountant and a poet mid-brainstorm.
Example
Set it near 0 to extract data cleanly; crank it up when you want ten weird names for a product.
How it actually works
It reshapes the probability distribution over the next token. High temperature flattens it so rare words get a chance; low temperature sharpens it so the safe pick almost always wins. Related knobs — top-p and top-k — trim the same distribution differently.
For product teams
A creativity dial — low for reliable extraction, high for brainstorming variety.
For engineers
Softmax scaling on next-token logits; higher flattens the distribution, lower sharpens it.
Related
- One of several “sampling” knobs, alongside top-p and top-k.
- Turn it down for extraction, up for ideation.
Read anything AI without the jargon
Look up any term in plain English, or save terms as you read with the free Chrome extension.
Open DecoderAdd to Chrome