OpenAI pricing spans two related but different evaluation paths: ChatGPT plans for employee use and API usage controls for software teams building with OpenAI models. For B2B buyers, that matters because the same vendor page set covers both workspace purchasing and application-level usage management, but the budgeting questions are different.
The ChatGPT pricing page covers Business and Enterprise plans, while the API documentation explains rate limits, usage limits, and throughput controls. That split affects software buyers, engineering leads, platform teams, and IT teams that need to budget for both end-user access and model usage inside products or internal tools. This guide covers those two levers — plan cost and the controls that govern API usage — rather than per-request token rates for individual models.
What it is and who it’s for
OpenAI API pricing is not a single line item. In practice, buyers are often evaluating a ChatGPT workspace for employees and API access for developers at the same time.
The ChatGPT pricing page in recent coverage covers Business and Enterprise plans. OpenAI describes Business as a secure workspace with company context and tools for teams, and says it includes access to ChatGPT and Codex across desktop and mobile apps. That package also includes centralized billing and administration, along with usage analytics, budgeting, and spend controls. OpenAI also says Business includes SAML SSO and MFA, has no training on your business data by default, is available for 2+ users, and is billed annually by default.
Enterprise is listed separately with custom pricing through sales. OpenAI says Enterprise includes an expanded context window for longer inputs and larger files, plus SCIM, EKM, user analytics, domain verification, role-based access controls, data residency in ten regions, 24/7 priority support, SLAs, custom legal terms, access to AI advisors for eligible customers, and invoicing, billing, and volume discounts.
The audience, then, splits into at least three buyer groups. Team leads may be evaluating ChatGPT Business as a managed end-user workspace. Enterprise IT and security teams may be focused on SAML SSO, MFA, SCIM, EKM, domain verification, role-based access controls, and data residency. Engineering and platform teams need the API-side view, where usage is controlled by model-specific rate limits, throughput measures, and monthly caps.
How it works
On the API side, OpenAI uses several rate-limit measures. The company documents RPM, RPD, TPM, TPD, IPM, and, for some streaming audio models, audio minutes per minute. In plain terms, that means usage is governed not only by how many requests an application sends, but also by model-specific throughput controls over time.
OpenAI says rate limits vary by model and are defined at the organization level and project level. It also notes that long context models such as GPT-5.5 can have a separate rate limit for long context requests, and that project-scoped token limit headers may appear when a project-scoped token limit applies.
Batch processing follows its own logic. OpenAI says Batch API queue limits are based on total input tokens queued for a given model, and pending batch jobs continue to count against that queue limit until they complete. For vector storage workflows, /vector_stores/{vector_store_id}/files and /vector_stores/{vector_store_id}/file_batches share a limit of 300 requests per minute for each vector store.
Usage limits are a different control from rate limits. OpenAI defines usage limits as monthly spend caps for the API, and says unsuccessful requests still contribute to per-minute limits.
Pricing and cost considerations
A practical budgeting approach is to treat the ChatGPT pricing page and the API rate-limits page as two separate inputs: one for workspace seat costs and one for API usage controls.
| Offering | Pricing model | Confirmed pricing/details |
|---|---|---|
| ChatGPT Business | Per user | $25.00 per user per month when billed monthly |
| ChatGPT Business | Eligibility and billing | Available for 2+ users and billed annually by default |
| ChatGPT Enterprise | Custom | Custom pricing through sales |
For API usage, OpenAI confirms the framework buyers need to budget against:
| API cost/control lever | How OpenAI describes it |
|---|---|
| Monthly spend cap | Usage limits are monthly spend caps for the API |
| Throughput controls | Rate limits vary by model and are set at the organization and project level |
| Batch usage | Queued input tokens count against Batch API queue limits until jobs complete |
| Failed calls | Unsuccessful requests still contribute to per-minute limits |
Published prices reflect OpenAI’s pricing page at the time of writing and can change, so treat the figures here as current rather than a fixed contract rate. For finance and platform teams, the practical implication is that a per-user ChatGPT plan does not answer the same budgeting question as API throughput and spend controls.
How to choose
The first decision is whether the need is employee productivity, product integration, or both.
If the immediate requirement is a managed AI workspace for staff, ChatGPT Business is the clearer fit in the provided material. OpenAI positions it as a secure team workspace with centralized billing, administration, analytics, budgeting, spend controls, access to ChatGPT and Codex across desktop and mobile apps, plus SAML SSO and MFA.
If the requirement is embedding OpenAI into software, internal systems, or customer-facing workflows, the API-side decision is less about a single list price and more about operational fit. Buyers comparing providers can set these rates against our Claude API pricing guide. Buyers should evaluate which models they expect to use, whether they need batching, and whether throughput will be constrained by requests, tokens, audio minutes, or queue depth.
A practical selection method is to map each use case to its main bottleneck. A lightweight assistant may be constrained by RPM. A document-heavy workflow may be constrained by TPM or long-context limits. A back-office processing pipeline may be constrained by Batch API queue limits.
If Enterprise is under consideration, the confirmed Enterprise-specific items on the pricing page include an expanded context window for longer inputs and larger files, SCIM, EKM, user analytics, domain verification, role-based access controls, data residency in ten regions, 24/7 priority support, SLAs, custom legal terms, access to AI advisors for eligible customers, and invoicing, billing, and volume discounts.
Limitations and gotchas
The biggest gotcha is that “pricing” does not tell the whole story for OpenAI deployments. Even when a plan or account appears affordable on paper, rate limits can shape what an application can actually do at launch.
Because those limits differ across the model lineup, a team cannot assume the same throughput on every model it tests. Another operational trap is retry behavior: unsuccessful requests still count toward per-minute limits, and OpenAI recommends exponential backoff and retry strategies for rate-limit errors.
Token settings also affect consumption. OpenAI says reducing max_tokens to match expected completion size can help control rate-limit consumption. It also says batching multiple tasks into one request can increase throughput when RPM is the bottleneck.
FAQ
Do these pages list per-token model prices? No. The ChatGPT pricing page covers plan pricing for Business and Enterprise, and the rate-limits page covers usage controls; neither lists per-request model token rates, so those have to be checked against OpenAI’s model pricing separately.
Which regions does Enterprise data residency cover? OpenAI lists ten: US, EU, UK, JP, CA, KR, SG, IN, AU, and UAE.
Is there a rate-limit measure beyond request and token counts? Yes. For some streaming audio models OpenAI applies an audio-minutes-per-minute limit, and long-context models such as GPT-5.5 can carry a separate limit for long-context requests.
