ChatGPT Enterprise Pricing (2026): What OpenAI Publishes, What It Doesn't, and When Business Is the Better Buy

The short answer

As of September 2026, OpenAI does not publish a ChatGPT Enterprise price. Its business pricing page says "Custom pricing" and routes you to a sales form. What it does publish, across the pricing page and its help center: Enterprise is an annual contract; usage is billed either from a credit pool or per token in dollars at rates set in your agreement; the per-token rate card is public (GPT-5.6 Sol at $4 in / $20 out per million tokens, the same as the API); and the plan next to it, ChatGPT Business, costs $20 per seat billed annually or $25 monthly for a Standard seat and $100 or $125 for a Premium seat, for teams of 2 to 200. I am not going to quote you a seat price from a procurement blog. This page is what OpenAI puts in writing, what Enterprise adds over Business, how the meter works, and why most companies under 200 people should be buying Business.

What OpenAI publishes about ChatGPT Enterprise pricing

All of this is from openai.com/business/pricing and OpenAI's help center, read on September 15, 2026.

• The price: "Custom pricing. Contact our sales team to discuss enterprise pricing." No seat price, no seat minimum, no contract minimum.
• The term: annual. The pricing FAQ says OpenAI offers "monthly plans for Go, Plus and Business and annual plans for Business and Enterprise." The token-billing article adds that monthly invoicing "does not make the agreement month-to-month."
• Two billing models: credit-based and token-based. Credit-based contracts buy a shared credit pool at the contract level. Token-based contracts, available on eligible agreements including ChatGPT for Financial Services, meter usage in tokens and bill in US dollars monthly in arrears, alongside any seat fees.
• Two seat types: standard ChatGPT seats, and Codex-only seats that OpenAI describes as usage-based "with no fixed cost per user per month."
• Commercial terms: invoicing, volume discounts, custom legal terms, SLAs, and a spending commitment that "can include" both ChatGPT Enterprise and the API. The help center's Business article lists what forces a contract: invoice billing, purchase orders, bank transfer, wire, ACH, net terms, split payments, Zero Data Retention, and BAAs.

The dollar figures live on ChatGPT Business pricing, the comparison for the rest of this post.

What Enterprise adds over Business

Here is what changes between the two plans on OpenAI's comparison table; everything not listed (SAML SSO, the admin console, admin roles, SOC 2 Type 2, domain verification, no training on your data, Codex, connectors, projects, deep research, unlimited everyday chat) is on both.

If none of the "No" rows is something a regulator, an IT policy, or a customer's security questionnaire is making you buy, Enterprise is a bigger context window and a faster queue at a negotiated price.

How the usage meter works: credits or tokens

The seat is not the price on Enterprise. This is the part most pricing guides skip, because OpenAI's help center is the only place it is written down.

Credit-based contracts. The workspace buys a shared pool of credits at the contract level. Everyday chat on the Instant model is unlimited and draws nothing. Everything else has a credit rate: a GPT-5.6 Sol message is 10 credits regardless of reasoning effort, a Sol Pro or GPT-6 Pro message is 50, agent mode is 30 per message, a deep research task is 50, an image is 5, voice is 1.25 credits a minute. ChatGPT Work and Codex meter per token instead: Sol is 100 credits per million input, 10 per million cached input, 500 per million output; Terra is 50 / 5 / 300; Luna is 5 / 0.5 / 30; GPT-6 Astra is 250 / 25 / 1,250. When the pool runs out, advanced features pause unless the workspace owner has enabled overages or bought more through the account team.

Token-based contracts. Everything, including Instant, is metered in tokens and billed in dollars, monthly in arrears, "at the rates in your agreement." The published card, per million tokens (input / cached / output): GPT-6 Astra $10 / $1 / $50; GPT-5.6 Sol $4 / $0.40 / $20; GPT-5.6 Terra $2 / $0.20 / $12; GPT-5.6 Luna $0.20 / $0.02 / $1.20. Deep research runs at $2 / $0.50 / $8, image generation at $8 / $2 / $30 per million image tokens, voice at $0.05 a minute plus the backend model's tokens, web search at $10 per 1,000 runs. Over 272K tokens of context, input doubles and output bills at 1.5x; fast mode is 2.5x; data residency adds 10%. Admins get a monthly workspace budget plus user and group limits.

One inference OpenAI does not state: 100 credits of Sol input on one card lines up with $4 on the other, about 4 cents a credit at list. That is my back-of-envelope, not a quote; the contract sets the real number.

What a seat actually costs on a token-based contract

Illustrations on the published token rates, with the seat fee left blank because OpenAI leaves it blank.

• A light seat. A few questions a day: 500,000 input tokens and 100,000 output on Sol a month is $2 plus $2. About $4 in usage, plus the seat fee.
• A working seat. Documents, spreadsheets, and long threads daily: 5 million in, 1 million out on Sol is $20 plus $20. About $40 in usage.
• A heavy Codex seat. A developer in Codex most of the day: 30 million in, 5 million out on Sol is $120 plus $100, about $220 before cached input pulls the first number down. OpenAI's own rate card says Codex "costs approximately $100 to $200 per developer per month" on average.

Now the same three people on Business. The light and working seats are $20 each on annual billing with usage included; the developer gets a Premium seat at $100 with 5x the usage and no 5-hour window. Three people, $140 a month, no meter to model, and credits available if the developer blows through Premium. That is why the Enterprise conversation for a small company almost always ends with "buy Business." What the 5-hour window actually caps is on ChatGPT usage limits.

Who actually needs ChatGPT Enterprise

A company under 200 people needs Enterprise when at least one of these is true, and each one is on OpenAI's own pages:

• Your identity provider has to provision and deprovision accounts automatically (SCIM). Business has SAML SSO for login but no SCIM.
• You need role-based access to decide which groups can use which models, tools, or connectors. OpenAI's help center notes GPT-6 Astra is off by default in Enterprise workspaces and enabled by role, a control Business does not offer.
• Security or legal wants the Compliance API feeding conversation logs into an archive or eDiscovery tool, or IP allowlisting, or Enterprise Key Management.
• A customer's vendor questionnaire asks for ISO 27001 on the plan itself, or a regulator requires data residency outside the US.
• You handle health data and need a BAA, or a client contract requires Zero Data Retention. Both are contracted-offering items per OpenAI's Business article.
• Accounts payable will not run a card and needs invoices, purchase orders, ACH, or net terms.
• You have more than 200 people, the top of the Business range.

If nothing on that list applies, you are a Business customer, and a sales call will not lower your price. The choice you actually face is Business versus a stack of Plus subscriptions on personal cards, which I walk through on ChatGPT Business vs Plus. Short version: Business at annual billing costs the same $20 as Plus and adds the workspace, the admin console, and the no-training default, so for any team of two or more it is the answer.

ChatGPT Business vs Enterprise for a 25-person company

Arithmetic on OpenAI's published seat prices. Twenty-five people: twenty-two who use ChatGPT for drafts, research, and spreadsheets, and three who live in it (an engineer in Codex, a marketer running images and deep research all day, an analyst on ChatGPT for Excel).

• Business: 22 Standard seats at $20 plus 3 Premium seats at $100, billed annually, is $740 a month, $8,880 a year. Add credits if the three heavy users exceed Premium's allowance; the spend controls cap that per seat type.
• Enterprise: 25 seats at a fee OpenAI does not publish, plus metered usage at the token or credit rates above, on an annual contract. The three heavy users cost the same usage on either plan; the difference is whether it is bundled into a Premium seat or itemized on an invoice.

Unless one of the seven triggers above applies, the twenty-five-person company buys Business, and the money it did not spend on a contract goes into the setup that makes twenty-five seats worth paying for: shared projects, a company-knowledge connector to the file server, and three workflows the team actually runs.

How to get a ChatGPT Enterprise quote, and what to ask for

The route is OpenAI's contact-sales form; select ChatGPT Enterprise as the product of interest and a representative follows up. Enterprise cannot be bought self-serve. Walk in with your headcount split between standard and Codex-only seats and a count of the heavy users, and ask for both a credit-based and a token-based quote, because the seat mix decides which is cheaper. Then get the numbers OpenAI does not print in writing: the per-seat fee, the seat minimum, the term, the price per credit or the discount off the token card, the overage policy, the true-up cadence, the data residency region, and the SLA. If an API commitment belongs in the same contract, say so up front. And name the specific control that sent you here, whether SCIM, the Compliance API, or the BAA, and confirm it in the contract. The pricing page marks features; the contract binds.

ChatGPT Enterprise vs Claude Enterprise on price

As of September 2026, Anthropic publishes the number and OpenAI does not. Claude Enterprise is listed at $20 per seat per month billed annually plus usage at API rates, with a self-serve checkout; the breakdown is on Claude Enterprise pricing. Both plans are annual, both meter usage on top of the seat, and both advertise the same control family: SCIM, role-based access, audit logging through a compliance API, data-retention controls, no training by default. For a company under 200 people the comparable pair is ChatGPT Business against Claude Team, priced identically at $20 a seat annual and $25 monthly.

The owner's answer

Most people who search "ChatGPT Enterprise pricing" run a company of five to a hundred people and typed the word because it is the biggest box on OpenAI's pricing page. If that is you: buy ChatGPT Business at $20 a seat, put the heavy users on Premium seats, turn on the spend controls, and spend the money you would have spent on a contract on setting the workspace up properly, which is what decides whether the seats pay for themselves. If a regulator, an IT policy, or a real audit requirement sent you here, Enterprise is the right product for the controls, and the metered usage above is the line to model before you sign an annual agreement whose seat price you had to ask for.