Most AI APIs charge by usage. Text models often price input tokens and output tokens separately, while image, audio, embeddings, search, or tool features may have their own pricing units.
In practice
The cost of one request depends on the model, prompt size, retrieved context, output length, and any extra features used. Long chats and large documents cost more because they send more tokens.
What to watch
Set budgets and monitor usage early. A small bug, retry loop, or oversized context can create unexpected cost.