Training
Prompt engineering
What is prompt engineering?
Prompt engineering shapes model behaviour through instructions supplied at request time. It is faster to iterate on than fine-tuning and costs tokens on every request; fine-tuning is the opposite trade.
It should always come first. If a prompt solves the problem, you have solved it in minutes with no training data, and the prompt also tells you what a fine-tune would need to learn.
The point to switch is usually one of three: the prompt has grown long enough to be a real per-request cost, it works most of the time but not reliably enough, or the behaviour you want cannot be described in words at all.
Related terms
Start with the free tier
A magic link creates your account, your tenant and your first API key. No card until you ask for compute.