Usage & billing
Billing is pay-as-you-go: only the tokens of text sent and received are deducted. There's no subscription fee and no minimum monthly volume.
What's a token
A token is a chunk of text, roughly 4 characters of English. A page of text is about 500–700 tokens. Both the request and the response are counted.
| Example | Approximate cost |
|---|---|
| A short question and chat reply | 300 – 800 tokens |
| Analyzing a page of a document | 1,500 – 3,000 tokens |
| An agent editing a code file | 5,000 – 30,000 tokens |
Why models cost differently
Each model has its own multiplier: heavier models burn through your balance faster, lighter ones slower. For the same amount of text, the deduction differs by exactly that multiplier. Values are shown next to each model in the catalog.
Where to see usage stats
- In your dashboard — usage by day, by model, and by key, so you can see how much each project consumes.
- Every response includes a
usageblock — handy to log on your side. - Access via a code: your own stats in the
/redeemdashboard, for your keys only.
How to spend less
- Don't send the entire conversation history — keep the last few messages plus a short summary.
- Route simple steps to a cheap model and hard ones to a strong model.
- Cap the response length with
max_tokens. - Cache responses to repeated questions on your side.
// keep the last few messages instead of the full history
const window = messages.slice(-8);Trimming request length usually saves more than switching models: you're mainly paying for the volume of text.