Two products, one sign-in

ChatGPT Business and the OpenAI Platform are two different products. You can sign in to both with the same email address, and they will still have separate access and separate billing. OpenAI is explicit about it: a Business subscription does not include API usage, and the two are billed through different systems.

ChatGPT Business is a product people use. The OpenAI Platform is a service software calls.- the distinction that resolves most questions

Everything else follows from that one sentence. A seat is a person; an API key is a program. One is bought per head at a fixed monthly price, the other is metered per token with no seats at all. They have different admin surfaces, different concepts for grouping people, and different invoices.

CHATGPT.COM

ChatGPT Business

A self-serve workspace plan that gives a team the ChatGPT app with company-level admin, shared settings, and business data terms.

Bought for
Employees who need to use AI directly - in a chat window, on files, in their working day.
PLATFORM.OPENAI.COM

OpenAI Platform

The developer product: API keys, models, usage dashboards, rate limits, and the projects that scope them.

Bought for
Software that needs to call a model - your app, your agent, your internal service, your pipeline.

ChatGPT Business

ChatGPT Business is the team tier of the ChatGPT app. You are buying the interface and the governance around it, not model capacity you can point at your own code.

What comes with it:

  • A workspace. The organizational unit. Members are invited into it, and the workspace - not each individual - holds the settings and the plan.
  • Roles. Workspace owners and admins manage members, seats, and billing; owners are the only ones who can change payment details and see invoices.
  • The app surfaces. Chat, file and document work, research, the workplace features, and Codex where the plan includes it - all inside ChatGPT.
  • Per-seat billing. A fixed monthly amount per member, managed under Workspace settings, plus workspace credits where a plan offers them.
  • Business data terms. The reason most companies move off individual Plus accounts: one agreement, admin visibility, and central control of who has access.

What it is not: a source of API keys. There is no key to copy out of a Business workspace, and nothing in the subscription that a developer can call from an application.

OpenAI Platform

The Platform is the developer side, at platform.openai.com. It exists to let software send work to a model and get a machine-readable result back - no browser, no chat window, no seats. This section is about the account you set up there; what the Platform does once you have a key (the Responses API, hosted tools, the agent loop, the Agents SDK) is the subject of What is the OpenAI Platform?.

What it gives you:

  • An organization, and projects inside it. Projects are how you separate work: each one has its own keys, members, models, and limits, so staging cannot spend production's budget.
  • API keys. The credential your code holds. Scoped to a project, revocable, and the thing you must never ship to a browser.
  • Usage-based billing. You pay per token consumed - by model, plus any tool or storage charges - either from prepaid credits or a payment method on file. There is no per-seat component.
  • Rate limits and spend controls. Per-project caps, usage dashboards, and limits that depend on your account tier rather than on any ChatGPT plan.
  • The dashboard. Logs, usage breakdowns, model availability, and the playground for trying a call before you write the code.

The cost model is the real difference in kind. A seat is predictable and flat; token spend is a function of traffic, so a Platform bill moves with what your product does that month. Budgeting one as if it behaves like the other is the classic first mistake - the same trap covered in what an agent costs.

Side by side

The full picture, one row per thing people actually get wrong.

ChatGPT Business OpenAI Platform
Website chatgpt.com platform.openai.com
Main purpose Employees use AI directly Developers build software on OpenAI models
Interface The ChatGPT app API, plus a dashboard around it
Typical use Chat, files, research, Codex, workplace AI Applications, agents, integrations, batch jobs
Grouping Workspace Organization + projects
People Workspace members, on seats Organization and project members - no seats
Admin Workspace owner and admins Organization and project owners and admins
Billing Per-seat subscription, plus workspace credits where offered Usage-based - per token, per model, plus tool and storage charges
API keys No Yes, scoped per project
API usage included No - not any amount Yes, and paid for separately

Note the two rows in the middle. "Workspace" and "Organization + projects" sound like synonyms and are not: they are separate account structures, with separate member lists and separate admins. Adding someone to the workspace does not give them access to the API organization, and vice versa.

What does not carry across

Almost every support question about these two products is really a question about what crosses the boundary. The short answer is: nothing does.

DOES NOT TRANSFER BETWEEN THE TWO
  • The subscription. Paying for Business seats buys no API usage - and there is no discount on token prices for having them.
  • Credits. Workspace credits stay in the workspace; API credits stay on the Platform. Neither converts into the other.
  • Rate limits. Platform rate limits and account tiers have nothing to do with your ChatGPT plan.
  • Membership. Workspace members are not organization members. Two invitations, two access lists.
  • Invoices and payment methods. Separate billing systems, separate settings pages, separate receipts for finance to reconcile.

The one thing that does carry across is your identity - the same email can sign in to both. That single shared detail is what makes people assume the rest is shared too.

Worth knowing if you also use Anthropic: the same seat-versus-token split exists there, but the boundary is deliberately porous rather than sealed - see Claude Team vs Claude Platform.

The practical consequence for finance: these are two line items, not one. Seats are a headcount cost that changes when you hire; token spend is a variable cost that changes when your product gets used. Approving them from the same budget without separating them is how a team ends up surprised by a bill that had nothing to do with the number of employees.

Which one you need

The test is who is doing the asking - a person, or a program.

  • People need to use AI at work. Buy ChatGPT Business. Seats, admin control, business data terms, nothing to build.
  • Software needs to call a model. Use the Platform. Keys, projects, usage billing, and code you own.
  • Both, which is the common case. Run them as two accounts on purpose: a workspace for staff and an organization with per-project keys for engineering. Expect two invoices and give each its own owner.

One more distinction worth holding on to, because it is the same shape of mistake: the model is not the product. GPT-class models appear inside ChatGPT, inside Codex, and behind the API, but those are different access routes with different terms and different bills - the same confusion covered in What is a model?. Buying access to a product is not buying access to the model underneath it.

References