01 / The Model
Input tokens
Everything sent to the model on a request: your message plus all the context around it. Cheaper than what comes back, because reading costs less than writing.
How you’ll hear it
“Why is a simple question still costing us?”
“We attach the whole policy manual to every question. That is all input.”
Avoid
Assuming only the question costs money. Everything sent with it counts.
The longer version
Input tokens are everything sent in a request: your question plus instructions, attached documents and earlier conversation. They are cheaper per token than output, but they add up quickly when large documents travel with every request.