What to know first
This rate card covers eligible Enterprise agreements with token-based billing. It does not establish the same billing model for Plus, Pro, or Business. Identify your plan and agreement before estimating usage costs.
Who should use the ChatGPT Enterprise token rate card?
Check the agreement, not just the plan name. Enterprise workspaces can have different billing models. If yours is moving from credits, confirm that the change has taken effect.
Under token-based Enterprise agreements, usage charges are separate from contracted seat fees. Monthly invoicing does not mean you can cancel monthly: review the contract term and any spending commitment.
| Your situation | What to check first |
|---|---|
| Choosing Plus, Pro, or Business | Plan features, subscription period, and checkout total |
| Buying or using Enterprise | Token or credit billing, plus any separate seat fees |
| Building with the API | API rates and the separate API bill |
Sources: OpenAI: Token-based billing for ChatGPT EnterpriseOpenAI: Managing billing for ChatGPT and the API platform
What are tokens, and why is message count not enough?
Tokens are units a model processes, not a fixed number of words or Chinese characters. Counts vary by language, model, and content. Input is supplied to the model, cached input reuses earlier input, and output is generated by the model.
For reasoning models, billed output can include internal reasoning you never see. A short answer can still consume substantial tokens. Use recorded usage and representative tasks when evaluating costs.
How to calculate token costs: a worked example
The rate card lists GPT-5.6 Sol at $4 for input, $0.40 for cached input, and $20 for output per million tokens. The usage below is hypothetical, not an average conversation.
Assume a task uses 10,000 uncached input tokens, 90,000 cached input tokens, and 2,000 output tokens. Count each category once; do not also charge cached tokens as uncached input.
| Category | Assumed tokens | USD / 1M tokens | Cost (USD) |
|---|---|---|---|
| Uncached input | 10,000 | 4.00 | 0.040 |
| Cached input | 90,000 | 0.40 | 0.036 |
| Output | 2,000 | 20.00 | 0.040 |
Category cost = tokens ÷ 1,000,000 × category rate
Total = 0.040 + 0.036 + 0.040 = USD 0.116
Sources: OpenAI: ChatGPT Rate Card (Enterprise token-based pricing)
What costs should you check beyond model tokens?
The rate card also covers voice, web search, image generation, fast mode, long context, and regional processing. Check the relevant units and model exceptions when estimating a total.
For each trial task, record model usage, features invoked, and special modes enabled. Compare the total cost of completing similar tasks before deciding whether a faster or more capable configuration is worth paying for.
Sources: OpenAI: ChatGPT Rate Card (Enterprise token-based pricing)
Does using Codex mean paying these Enterprise rates?
It depends on your access method and agreement. Included subscription usage, additional credits, Enterprise usage billed in USD, and API-key billing need separate checks. API-key usage is charged at applicable API rates rather than following a subscription allowance.
Before comparing prices, confirm your sign-in method, workspace, and the billing units on your usage page. Sharing a model name does not make two accounts' billing arrangements identical.
Sources: OpenAI: Codex pricing and usage
How can an enterprise control its usage budget?
Token-based Enterprise billing supports a monthly workspace budget in USD and user and group limits. User limits and the workspace budget apply independently; raising one person's limit does not raise the workspace budget.
Usage alerts notify you; they do not stop usage. A task already running can finish slightly above a user's limit, so review alerts, limits, and invoices separately.
- Run representative tasks and record their costs and the quality of the results.
- Estimate a baseline from the number of users and task frequency, allowing room for experiments and rework.
- Assign responsibility for limit changes and increase requests, then review costly tasks regularly.
Sources: OpenAI: Manage usage limits and overages in ChatGPT Enterprise and Edu
Four checks before subscribing or upgrading
AISub compares regional reference prices for ChatGPT web subscriptions. Use it to compare subscription prices; estimate Enterprise token usage separately under your agreement.
- Access: do you need a ChatGPT web subscription, an enterprise workspace, or API development access?
- Total cost: will you need additional usage beyond a subscription or seat fee?
- Tasks: test the plan against your own writing, analysis, or coding work.
- Payment terms: check the billing period, currency, taxes, and contract before buying.
Frequently asked questions
Does a ChatGPT subscription include free API usage?
No. ChatGPT and the API platform manage billing separately. For paid API access, check your payment setup and usage on the API platform.
Sources: OpenAI: Managing billing for ChatGPT and the API platform
Can I convert my old enterprise credit total directly into a dollar budget?
No. A previous credit total is not the dollar cost of the same activity under token-based billing. Recalculate using actual usage and your agreement.
Why can a short answer consume many tokens?
Visible text does not represent all usage. Input context and internal reasoning affect consumption. Check usage records rather than counting the words in an answer.
Official sources and review notes
Sources checked on September 28, 2026. Rates and eligibility rules can change; confirm the latest official guidance and your own agreement.
- OpenAI: ChatGPT Rate Card (Enterprise token-based pricing)
- OpenAI: Token-based billing for ChatGPT Enterprise
- OpenAI: Understanding and counting tokens
- OpenAI: Managing billing for ChatGPT and the API platform
- OpenAI: Codex pricing and usage
- OpenAI: Manage usage limits and overages in ChatGPT Enterprise and Edu
AISub prepared this guide from public OpenAI documentation, with an independent worked example and purchasing advice. AISub is an independent price index and is not affiliated with OpenAI.