Pay-as-you-go prices and terms
Optional additional AI usage for eligible Pro subscriptions. Enabling it requires your explicit agreement and a spending limit. Availability remains controlled while billing validation is completed.
Terms version payg-v1 · Price schedule reviewed 2026-10-05
Your billing choice
Additional charges begin only after your plan’s included AI allowance is exhausted and you have enabled pay-as-you-go. Your spending limit applies to additional AI charges in each subscription period, resets at renewal, and does not carry over. It is separate from your subscription fee.
Pay-as-you-go stays enabled at renewal until you turn it off. Turning it off prevents future additional usage; charges already incurred remain payable. Changes take effect after current AI requests finish. We may stop a new request before your spending limit if there is insufficient room for its possible usage. Free accounts must upgrade to an eligible subscription before enabling pay-as-you-go.
Text usage prices
All prices below are final customer prices in USD per 1,000,000 tokens for supported Standard text processing. Tokens are pieces of text used by the model. Input includes the prompt, document content, conversation history, and tool results sent to the model. Output includes generated reasoning as well as visible responses. A task can involve multiple model calls, and all model work performed for the task is counted.
| Model | Input | Cached input | Cache write | Output |
|---|---|---|---|---|
| gpt-4.1 | $2.60 | $0.65 | Not separately priced | $10.40 |
| gpt-5.2 | $2.275 | $0.2275 | Not separately priced | $18.20 |
| gpt-5.4 | $3.25 | $0.325 | Not separately priced | $19.50 |
| gpt-5.5 | $6.50 | $0.65 | Not separately priced | $39.00 |
| gpt-5.6-luna | $0.26 | $0.026 | $0.325 | $1.56 |
| gpt-5.6-terra | $2.60 | $0.26 | $3.25 | $15.60 |
| gpt-5.6-sol | $5.20 | $0.52 | $6.50 | $26.00 |
| gpt-6-sol | $2.60 | $0.26 | $3.25 | $13.00 |
Input tokens are charged in their applicable category: ordinary input, cached input, or cache write. A cache write is not also charged as ordinary input. “Not separately priced” means the model has no separate cache-write rate.
Requests above 272,000 input tokens
For the models in this table, a request with more than 272,000 input tokens uses these prices for the entire request, including its output. GPT-4.1 and GPT-5.2 retain the prices in the first table.
| Model | Input | Cached input | Cache write | Output |
|---|---|---|---|---|
| gpt-5.4 | $6.50 | $0.65 | Not separately priced | $29.25 |
| gpt-5.5 | $13.00 | $1.30 | Not separately priced | $58.50 |
| gpt-5.6-luna | $0.52 | $0.052 | $0.65 | $2.34 |
| gpt-5.6-terra | $5.20 | $0.52 | $6.50 | $23.40 |
| gpt-5.6-sol | $10.40 | $1.04 | $13.00 | $39.00 |
| gpt-6-sol | $5.20 | $0.52 | $6.50 | $19.50 |
What can be billed
The initial offering covers supported text model requests within enforced request limits. Local document tools have no separate tool-call price in this schedule, but model calls and text sent around those tools count. Hosted search, media generation, and other separately priced services are outside this initial offering. This schedule does not promise those services or other processing tiers.
Cancelling a task does not remove charges for model work already performed. We charge for recorded usage, not the task’s advance estimate. Usage is accumulated before invoice rounding to USD cents. Applicable taxes, if any, are separate. New prices require an updated schedule and billing terms before they apply.