Prompt
The input you give a model — the text that sets up what you want it to do.
Think of it like
The brief you hand a freelancer: how good the work is depends a lot on how clear you were.
Example
“Summarize this email in two sentences” plus the email is the prompt; the summary is the completion.
How it actually works
The prompt is everything you feed in: instructions, context, examples, the actual question. The model conditions its entire response on it. Because behavior is so sensitive to wording, prompting is a real skill — though robust models need less coaxing than they used to.
For product teams
Often the cheapest lever you have — rewording the prompt beats fine-tuning for many problems.
For engineers
The input token sequence conditioning generation; includes system, user, and any in-context examples.
Related
- Completion — The model replies with a completion.
- Prompt engineering — Crafting it well is prompt engineering.
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