Pricing and Billing

Inference is billed by the tokens you send and the tokens the model generates. Rates vary by model. Check the Models page for current prices.

Inference pricing

Prices are listed per million tokens:

• Input: your prompt, system instructions, and conversation history sent with the request.

• Output: the tokens generated by the model.

• Cached input: input tokens read from the prompt cache, billed at a lower rate when the model offers one.

Repeated content, such as a system prompt, may benefit from caching. The cached rate applies only to tokens actually read from the cache.

You pay for the tokens used, not a full million. For example, at an input rate of $1 per million tokens, 1,000 input tokens cost $0.001. Output tokens are charged separately at the model's output rate.

Track your usage and costs

View your usage to see recorded inference costs, input and output tokens, and cached tokens. Filter by model, API key, or period to understand which activity contributes to your spending. See Understand your usage for a walkthrough.

Paying for usage

Paid inference usage is deducted from your prepaid balance. See the Billing guide to add funds or set up automatic top-up.