Model Serving & Economics
Token cost
The price charged for the amount of text or tokens processed or generated by a model.
Example
A long prompt costs more because the provider charges for each processed token.
Why it matters
It links charges to how much material the service handles rather than just counting visits.
You'll hear it
“That long conversation used more of our budget than expected.”
Can you control it?
Sometimes
Sometimes. Some products or workflows let you influence token cost, while others keep it fixed or manage it automatically. Check the settings and documentation for the tool you are using.
Common questions
- Is a token the same as a word?
- No. A word can be split into several pieces, and punctuation can count too.
- Can the same text cost different amounts with different models?
- Yes. Models can split text differently and providers can charge different rates.
- Does a failed request always cost nothing?
- No. Billing can depend on how much processing occurred before the failure; check the provider's rules.