Learn
ChatGPT Enterprise Pricing: What OpenAI Publishes and What It Does Not
OpenAI publishes no ChatGPT Enterprise seat price. It does publish how usage is billed: credit-based and token-based rate cards, such as GPT-6 at $2/$10 per million tokens. See what each agreement charges and what to ask sales for.
OpenAI does not publish a price for ChatGPT Enterprise. Its pricing page lists Enterprise as custom pricing through sales, with credit-based and token-based agreements. What OpenAI does publish is how usage is billed under each agreement: a credit rate card, for example 50 credits per Deep Research task, and a token rate card in US dollars, for example $2 per million input tokens and $10 per million output tokens for GPT-6 in Chat. Seat prices, minimum seat counts, the dollar value of a credit and overage rates are set in each contract. ChatGPT Business is the self-serve alternative at $25 per seat a month, or $20 billed annually.
What OpenAI publishes about ChatGPT Enterprise pricing
Pricing element | What OpenAI publishes |
|---|---|
Seat price | Not published; Enterprise is custom-priced through OpenAI sales |
Pricing models | Credit-based agreements and token-based agreements |
Minimum seats and contract term | Not published; Enterprise is sold on annual plans |
Usage on credit-based agreements | The workspace buys a shared credit pool at the contract level, used at the published credit rate card |
Usage on token-based agreements | Billed in US dollars at the published Enterprise rate card, plus any seat fees in the agreement |
Dollar price of a credit | Not published |
Per-user limits | None by default on credit-based pricing; admins can set spend controls by group |
Overage | Owners can enable overages or buy more credits through the account team; overage rates are not published |
Adding users mid-term | Qualifying added seats are generally reconciled through quarterly true-ups on seat-bundled credit contracts |
Discounts | Volume discounts; up to 75% off for nonprofits through OpenAI for Nonprofits |
Billing | Invoicing available through sales |
Source: OpenAI's ChatGPT pricing page and its help articles on ChatGPT Enterprise, flexible pricing and the Enterprise rate cards, checked October 11, 2026.
Token-based Enterprise rates
On token-based agreements, usage across Chat, ChatGPT Work and Codex is charged per million tokens, and GPT-6 Instant usage is billable too. These are OpenAI's published standard rates; your agreement can apply discounts.
Model or feature | Where it is used | Input | Cached input | Output |
|---|---|---|---|---|
GPT-6 (GPT-6 Sol) | Chat | $2.00 | $0.20 | $10.00 |
GPT-6 Pro (GPT-6 Astra) | Chat | $10.00 | $1.00 | $50.00 |
GPT-6.1 Sol | Work and Codex | $2.00 | $0.10 | $10.00 |
GPT-6 Astra | Work and Codex | $10.00 | $1.00 | $50.00 |
GPT-6 Luna | Work and Codex | $0.10 | $0.01 | $0.50 |
Prices in US dollars per million tokens, from OpenAI's ChatGPT Enterprise token-based rate card, checked October 11, 2026. Additional charges apply for Voice ($0.05 per minute plus the backend model's tokens), web search ($10 per 1,000 web runs plus content tokens), long-context input above 272K tokens in Work and Codex (twice the input rate and 1.5 times the output rate, with no surcharge for GPT-6 Astra in Codex) and regional data processing (1.1 times the standard rate).
Credit-based Enterprise rates
On credit-based agreements, usage draws from the workspace's credit pool. OpenAI publishes the credit cost of each feature but not what a credit costs in dollars; that is set in the agreement.
Usage | Credits |
|---|---|
GPT-6 message in Chat (Medium, High or Extra High) | 10 per message |
GPT-6 Pro message in Chat | 50 per message |
Instant | Unlimited, subject to guardrails |
Agent mode | 30 per message |
Deep Research | 50 per task |
Image generation | 5 per generation |
Voice | 1.25 per minute |
Source: OpenAI's credit-based rate card for Business, Enterprise and Edu, checked October 11, 2026. Work and Codex usage on credit-based agreements is charged in credits per million tokens instead, at rates OpenAI lists in the same article. OpenAI says Codex costs approximately $100 to $200 per developer per month on average, with large variation by model, concurrency, automations and fast mode.
What the published numbers do not tell you
The rate cards set the price of usage, not the size of the bill. Without a published seat price, minimum or credit value, the only way to know what Enterprise will cost your organization is a proposal from OpenAI. Before comparing a proposal with Business, ask for these terms in writing:
- Seat fee and seat types. Whether standard ChatGPT seats carry a fixed fee and whether Codex-only seats, which are usage-based on credit agreements, are offered.
- Minimum seats and term. The seat commitment and the length of the annual agreement.
- Which rate card applies. Credit-based or token-based, and any discount on the published rates.
- Credit value and expiry. On credit agreements, what a credit costs and when unused credits expire; OpenAI says the Order Form defines both.
- Overage terms. Whether overages are allowed once the pool runs out, at what rate, and with what alerts.
- True-up timing. How added seats are billed, usually through quarterly true-ups on seat-bundled credit contracts.
When Business is enough
ChatGPT Business is self-serve: Standard seats cost $25 a month or $20 billed annually, Premium seats $125 or $100, with a two-seat minimum and card billing. It covers SSO, central billing and spend controls. Enterprise is the route when you need SCIM, enterprise key management, custom data retention, data residency in its listed regions, invoicing or contractual terms such as SLAs. ChatGPT Business vs Enterprise compares the two plans' controls in detail, and ChatGPT Plus vs Business covers the step up from a personal plan.
The OpenAI API is a separate product with its own pricing. OpenAI manages ChatGPT Enterprise workspaces and API platform organizations separately, so membership in one does not grant access to the other.
Evidence boundary
Official sources
Editorial guidance grounded in official product sources.
FAQ
Common questions
How much does ChatGPT Enterprise cost?
OpenAI does not publish a price. Enterprise is custom-priced through OpenAI sales, with credit-based or token-based agreements, and the seat fee, minimum seats and term are set in each contract. OpenAI does publish its usage rate cards, such as GPT-6 at $2 per million input tokens and $10 per million output tokens on token-based agreements.
Is there a minimum number of seats for ChatGPT Enterprise?
OpenAI does not publish an Enterprise seat minimum; it is set in the agreement. The self-serve ChatGPT Business plan needs at least two seats and costs $25 per seat a month, or $20 billed annually.
What is the difference between credit-based and token-based ChatGPT Enterprise pricing?
On credit-based agreements, the workspace buys a shared credit pool and each feature costs a published number of credits, such as 50 per Deep Research task. On token-based agreements, usage is charged in US dollars per million tokens at OpenAI’s published Enterprise rates, plus any seat fees in the agreement.
How much does ChatGPT Enterprise usage cost per token?
On token-based agreements, OpenAI lists GPT-6 in Chat at $2 per million input tokens and $10 per million output tokens, GPT-6 Pro at $10 and $50, GPT-6.1 Sol in Work and Codex at $2 and $10, and GPT-6 Luna at $0.10 and $0.50. Agreements can apply discounts to these rates.
Is ChatGPT Enterprise usage unlimited?
Not in general. On credit-based agreements, Instant chat is unlimited subject to guardrails, while other features draw on the shared credit pool and pause when it runs out unless overages are enabled. On token-based agreements, Instant usage is billable at the published rates too.
Do nonprofits get a discount on ChatGPT Enterprise?
Yes. OpenAI says nonprofits can get up to 75% off ChatGPT Business or ChatGPT Enterprise through OpenAI for Nonprofits, by contacting its sales team.
Next steps
Take the next buying step
Use these next pages to confirm the plan, tool, or alternate route that fits once the spend boundary is clear.