Inference
1 min read · Reviewed 17 Oct 2026
In short
The moment a model actually produces an answer — and what you pay for.
In plain English
Training happens once and costs a fortune. Inference is what happens every time you press send: the model runs and produces output. It is a separate cost, paid per use.
A real example
A free chat tool is absorbing your inference costs to acquire you. An API bill is the same costs, itemised. This is why usage limits appear on free tiers, and why 'unlimited AI' offers tend to acquire caveats.
Why it matters
It explains the pricing of every AI product you will meet. If something is genuinely unlimited and free, either the model is small, the limits are hidden, or you are the product.
