Reading an API bill without panicking

What do the lines on my LLM API bill mean?

Three notions are enough to decode it all: tokens, context and model. The first bill is frightening less for its amount than for how opaque it is; once you understand those three words, you know exactly where every euro goes and where to start acting.

Why is the token the unit that counts?

Because everything is billed by the token, on the way in as much as on the way out, and the input is everything you send to the model on every call, including the context you carry around without reading it. That is often where, in that useless context repeated on every request, the bill balloons without usage going up.

Why does the choice of model multiply the bill?

Because at equal volume, two models can show a cost that differs by a factor of ten. Spotting which model handles what means spotting the savings right away: a simple task handed to the most expensive model is money thrown away.

A bill you understand stops being a source of dread. It becomes a lever.