ChatGPT Enterprise Pricing in 2026: What Is Published, and What Is Not
ChatGPT Business costs $20 per seat a month on an annual plan. Enterprise is custom. How to budget for both, and why credits and tokens change the math.
ChatGPT is often the AI tool employees already use before IT buys it. When the company formalizes it, finance needs a number. OpenAI publishes some of that number and keeps the rest for the sales conversation.
What OpenAI publishes (October 2026)
| Plan | Price per seat per month | Billing | Notes |
|---|---|---|---|
| ChatGPT Business, Standard seat | $20 | Billed annually | For teams of 2 to 200 people |
| ChatGPT Business, Standard seat | $25 | Billed monthly | Same seat, no annual commitment |
| ChatGPT Business, Premium seat | $100 | Billed annually | Higher limits |
| ChatGPT Business, Premium seat | $125 | Billed monthly | Higher limits |
| ChatGPT Enterprise | Custom | Annual contract | Contact OpenAI sales |
OpenAI's pricing page lists ChatGPT Enterprise as custom pricing. It also says Enterprise and Business customers can buy credits for more access, and that Enterprise offers credit-based and token-based pricing.
What is not published
- The Enterprise price per seat. It is negotiated, so two companies of the same size can pay different amounts.
- The seat or contract minimum for Enterprise.
- The price of credits and the size of credit pools.
- Volume discounts and overage terms.
Budgeting: a simple model
For a team on ChatGPT Business Standard seats, the published prices make the math easy. 200 people on annual billing cost 200 × $20 × 12 = $48,000 a year. The same 200 seats on monthly billing cost 200 × $25 × 12 = $60,000, so the annual plan saves $12,000, or 20%.
For Enterprise, model three numbers before you negotiate: the seats you need on day one, the seats you expect in 12 months, and how many of today's users are actually active. The last number is your leverage. Paying for seats nobody uses is the most common overspend in AI licensing.
Why credits and tokens change the budget
Seat pricing is predictable: the bill only moves when headcount moves. Credit-based and token-based pricing behave like metered cloud spend: the bill moves with usage, heavier models and longer conversations. A team that discovers a new workflow can double its usage in a month. That is good for adoption and hard for a fixed budget.
- Treat credits and tokens as a separate budget line from seats, with its own owner.
- Watch usage per team weekly in the first months of a contract, when habits form.
- Alert owners at 80% of a monthly allowance, before overages start.
How KeepAir helps
KeepAir puts ChatGPT seats and usage next to every other AI vendor, from Microsoft 365 Copilot to model APIs. Every seat maps to a person and team, idle seats are flagged with their cost, and usage-based spend gets budgets and alerts like any other line. When renewal comes, you negotiate with real utilization numbers, not estimates.