Credits pay for the models that answer your questions — ChatGPT, DeepSeek and the rest, called through their API. You pay for what you use, and nothing else.
| Model tier | $10 | $25 | $60 |
|---|---|---|---|
| Fasteveryday questions | 375 | 1,127 | 3,007 |
| Balancedharder reasoning | 126 | 378 | 1,010 |
| Most capablethe flagship models | 27 | 83 | 222 |
A question is one round of asking across your documents. Research plans run many of them, so they cost more — and you are charged for what a run actually uses, never the estimate.