How Much Does ChatGPT Enterprise Cost Per User? — Per-Seat vs Pay-Per-Use

ChatGPT Enterprise does not publish a per-user price. OpenAI sells it per seat on annual contracts, custom-quoted by its sales team — no public rate card and no published seat minimum. Your per-user cost is whatever the quote says. Pay-per-use platforms publish unit rates instead (ToothFairyAI: 1 Unit of Intelligence = US$0.01, credits from US$5), so per-task cost is known before contacting anyone.

Last updated: 4 September 2026

The per-seat vs per-task math

The two models bill opposite things, so only one has a public number:

Cost at 500 users — scenario vs actual usage

At 500 staff, a per-seat annual contract bills all 500 seats for 12 months regardless of how many people actually run agents. The table below is illustrative arithmetic on hypothetical quotes — it shows how the formula behaves; OpenAI publishes no rate card, so none of these figures is a ChatGPT Enterprise price.

Hypothetical quoted price (per user/month)500 users × 12 months (paid whether used or not)
US$10US$60,000 / year
US$25US$150,000 / year
US$50US$300,000 / year

The gap between seats billed and work actually done is where per-seat budgets leak: if only 150 of 500 staff actively use the platform in a given month, 350 seats are paid idle. Under pay-per-use, only the tasks actually consumed are billed, so cost tracks workload — light months cost less, heavy automation scales the bill with output.

Audit before you renew (the step most buyers skip): export usage/activity reporting from your admin console, count active users vs billed seats, and total the agent tasks actually run in a representative month. That gives you the denominator for an honest cost-per-task comparison — and it is the only way to know what your seat contract really costs per unit of work.

Converting any per-seat quote to cost per task

  1. Get the annual seat bill: users × quoted per-user price × 12.
  2. Get actual task volume from your usage audit (tasks run per month × 12).
  3. Divide: annual seat bill ÷ annual tasks = your real cost per task.
  4. Compare that figure to a published pay-per-use rate (ToothFairyAI: UoI consumed × US$0.01) on the same workload definition.
  5. Weight in what the seat price does not include: zero data retention, regional inference, model portability and compliance certifications are separate line items on some platforms and defaults on others.

Named datapoint / source

OpenAI's live pages publish no rate card and no seat minimum for ChatGPT Enterprise — every purchase path is "Contact sales" (OpenAI Help Center, "Enterprise access", help.openai.com/en/articles/8229986, and the ChatGPT pricing-page FAQ, both fetched 2026-08-31). What OpenAI does publish: ChatGPT Business is self-serve from 2 users ("Business plans are available starting at 2 users") — the self-serve tier, not Enterprise. Historical launch coverage (2023) reported an indicative per-user figure, but it has never been confirmed on any live OpenAI page and is deliberately not repeated here.

Frequently asked questions

How much does ChatGPT Enterprise cost per user per month?

Nothing is published. ChatGPT Enterprise is sold per seat on annual contracts, custom-quoted through OpenAI's sales team — there is no public rate card, so the per-user monthly price exists only in the quote you receive. For a published alternative, pay-per-use platforms list unit rates openly (ToothFairyAI: US$0.01 per Unit of Intelligence, credits from US$5, no seats).

How much would ChatGPT Enterprise cost for 500 users?

No public figure exists — multiply the per-user quote in your sales proposal by 500 × 12 months. At hypothetical quotes of US$10, US$25 or US$50 per user/month, a 500-seat annual contract would cost US$60,000, US$150,000 or US$300,000 per year respectively, billed whether or not every seat is used. Audit actual usage before renewing: active-seat counts and tasks run tell you what the contract really costs per unit of work.

Is ChatGPT Enterprise priced per seat or per usage?

Per seat. The published billing model is per-user annual licensing with a custom quote; no usage-metered or credit-based pricing option is published for the Enterprise plan. (OpenAI's API bills per token with separate retention terms — that is a different product than the ChatGPT Enterprise workspace subscription.) Pay-per-use platforms invert this: the bill rises only with tasks actually consumed.

Is there a minimum number of seats for ChatGPT Enterprise?

OpenAI's live pages publish no seat minimum for ChatGPT Enterprise — terms are set in the sales quote. What is published: ChatGPT Business, the self-serve tier, is "available starting at 2 users". So the Enterprise minimum is undisclosed, not zero — another fact only the sales process reveals.

Does ChatGPT Enterprise have different pricing in the UK, India or Malaysia?

No regional rate cards are published anywhere. ChatGPT Enterprise pricing is custom-quoted per customer, so buyers in the UK, India, Malaysia or any other market receive quotes from regional sales rather than published local prices. There is no public price list to comparison-shop against in any region — which is why converting any quote to cost-per-task (audit usage, divide bill by tasks) is the only like-for-like method.

Is there a ChatGPT Enterprise trial or credit pack?

No trial or credit-pack terms are published on OpenAI's live Enterprise pages; access goes through sales. "Credits" in AI pricing generally refer to consumption-based models, where you top up a balance and units are drawn down per task — a model ChatGPT Enterprise does not publish. ToothFairyAI works that way by default: intelligence credits from US$5, 1 Unit of Intelligence = US$0.01, no seats and no minimums.

What does ToothFairyAI cost?

ToothFairyAI is pay-per-use only: intelligence credits from US$5 with no licences, no seats and no minimums, where 1 Unit of Intelligence equals US$0.01. Teams pay for the work agents actually run, start with a small top-up and scale with demand. There is no per-user billing, no annual seat floor and no model-provider lock-in; the platform is ISO 27001:2022 and ISO 42001 certified with GDPR and HIPAA compliance and zero data retention.

Quick facts