Quick Answer
ChatGPT Enterprise has no public list price. Reported 2026 contracts average about $60 per user per month within a $45 to $75 range, with 150-seat minimums and annual prepayment, putting the floor near $108,000 a year. Newer eligible agreements add usage-based billing under OpenAI's published token rate card.
On June 21, 2026, Samsung Electronics switched on ChatGPT Enterprise and Codex for its entire South Korean workforce and every employee in its global Device eXperience division, tens of thousands of people across multiple continents. OpenAI described it as one of its largest enterprise deployments to date.
The reversal took three years. In May 2023, Samsung banned generative AI tools across company devices and internal networks after engineers pasted proprietary source code into the public version of ChatGPT. Nvidia hit a comparable milestone earlier in 2026, putting Codex in front of more than 10,000 employees across engineering and non-engineering roles.
Both announcements share a shape: headcounts, and no prices. Enterprise AI deals get celebrated in seats and almost never in dollars, so a finance leader pricing a rollout is left assembling the number from sales calls, buyer-reported contract data, and one document OpenAI actually does publish.
That document rewrote how enterprise ChatGPT gets budgeted, and it is where any 2026 budget has to start.
How much does ChatGPT Enterprise cost?
OpenAI doesn’t publish ChatGPT Enterprise pricing. The plan, OpenAI’s top business tier for large organizations, is quote-based, negotiated per organization, and billed as a prepaid annual invoice with no month-to-month option. Widely reported 2026 figures cluster around $60 per user per month, a typical range of $45 to $75, and a 150-seat minimum. OpenAI has never confirmed any of them.
Run the reported numbers and the practical ChatGPT Enterprise entry point lands around $108,000 a year: 150 seats at roughly $60 per user per month, paid up front. The same 150-seat floor shows up across the public trackers, and teams below it usually get steered to Business, which starts at two seats.
| Component | What OpenAI publishes | What buyers and pricing trackers report for 2026 |
|---|---|---|
| Per-seat price | Nothing (“contact sales”) | About $60/user/month average, $45 to $75 range |
| Seat minimum | Nothing | 150 seats, consistently reported |
| Contract term | Billed annually | Annual, prepaid, no monthly option |
| Usage billing | Token rate card (public) | Included usage and overage terms negotiated per deal |
| Discounts | Up to 75% off for nonprofits | Volume and multi-year discounts negotiated |
Scale moves the number. One licensing advisory’s mid-2026 benchmarks report $55 to $60 per seat at 150 to 999 seats, $48 to $55 in the 1,000 to 4,999 band, and $40 to $48 above 5,000 seats. Treat all of it as a negotiating baseline: what you actually pay depends on seat count, committed spend, and discount terms, and OpenAI states plainly that the signed agreement controls the rates.
There’s a reason enterprise pricing has become harder to pin down: the price stopped being one number. On eligible token-based agreements, the per-seat fee is only part of the bill. The rest is metered.
The ChatGPT Enterprise rate card is OpenAI’s published price list for usage-based billing on eligible token-based agreements. It applies to eligible Enterprise agreements billed in USD by token consumption, and it prices every surface of the product: Chat, ChatGPT Work, Codex, voice, web search, images, and the Excel, PowerPoint, and Workspace Agents tools.
The mechanics mirror API billing. Every request is metered as input tokens, cached input tokens, and output tokens, each billed at the model’s rate per million. If tokens are new territory, CloudZero’s token-based pricing explainer covers how the units work.
Here are selected rates from OpenAI’s Enterprise rate card, in USD per 1M tokens:
| Model | Input | Cached input | Output |
|---|---|---|---|
| GPT-6 Astra (Work/Codex) | $10.00 | $1.00 | $50.00 |
| GPT-5.6 Sol | $4.00 | $0.40 | $20.00 |
| GPT-5.6 Terra | $2.00 | $0.20 | $12.00 |
| GPT-5.6 Luna | $0.20 | $0.02 | $1.20 |
| GPT-5.5 Pro (Chat) | $30.00 | N/A | $180.00 |
List prices above are from OpenAI’s first-party Enterprise rate card, checked September 30, 2026. Partner clouds and credit-billed workspaces publish different figures for the same models.
Feature charges stack on top of model tokens. Voice runs $0.05 per connected minute plus backend model tokens. Web search costs $10 per 1,000 runs plus content tokens. Deep research bills at $2 input and $8 output per million on its dedicated model. Image generation runs $8 input and $30 output per million tokens.
Then come the multipliers, and they’re where budgets break:
| Multiplier | Rate impact |
|---|---|
| Long context (over 272K input tokens) | 2x input, 2x cached input, 1.5x output |
| Fast mode (Codex and Work only) | 2x on USD token rates (GPT-6 Astra Ultrafast: 6x) |
| Regional processing (data residency) | 1.1x on everything |
A Codex task on GPT-5.6 Sol that consumes 2,000 input and 800 output tokens costs about 2.4 cents. Run the same task in fast mode, OpenAI’s higher-priority processing tier, with regional processing enabled and it costs roughly 5.3 cents. Nothing about the work changed. The configuration did, and the bill grew 120%.
Data residency deserves its own line item: OpenAI offers residency in ten regions for eligible Enterprise customers, from the US and Europe to Japan and the UAE, and on token-billed agreements that compliance choice carries the 1.1x meter everywhere it applies.
Three fine-print items matter before any forecast:
- The card covers eligible token-based Enterprise agreements. Enterprise workspaces whose agreements have not moved stay on the separate credit-based rate card, and token-based billing is not offered at all for ChatGPT Edu, Healthcare, or Business. OpenAI directs unsure customers to their account team.
- GPT-5.6 Sol’s rates are promotional for token-based Work and Codex usage through at least November 21, 2026. Budgets built on today’s Sol pricing carry a repricing date.
- ChatGPT for PowerPoint was free for Enterprise customers through August 6, 2026. It now bills at the same token rates as the Excel tool.
The shift didn’t arrive overnight. OpenAI moved Codex from per-message credits to token metering on April 2, 2026, and extended it to all existing Enterprise plans, including Edu, Health, Gov, and ChatGPT for Teachers, on April 23. Each step moved ChatGPT further from a seat product and closer to a metered platform, and the Codex pricing guide tracks the piece of it engineering teams feel first.
ChatGPT Business vs Enterprise: what’s the difference?
Business is the self-serve team plan; Enterprise is the negotiated contract. Most buyers choosing between Business and Enterprise are really testing three forcing functions: compliance requirements, identity infrastructure, and scale.
Official ChatGPT Business pricing is $20 per seat per month billed annually, or $25 monthly, with a two-seat minimum. Everything in the Enterprise column below is negotiated:
| ChatGPT Business | ChatGPT Enterprise | |
|---|---|---|
| Price | $20/seat/month annual, $25 monthly | Custom, negotiated |
| Minimum | 2 seats | Not published; 150 seats commonly reported |
| Purchase | Self-serve | Sales-negotiated, annual contract |
| Identity | SAML SSO | SSO plus SCIM provisioning |
| Data | No-training guarantee | No-training guarantee, plus data residency options |
| Compliance | SOC 2 Type 2, ISO 27001 family | HIPAA BAA eligibility, expanded admin controls |
| Usage billing | Credit-based flexible pricing | Credit-based or token-based, per agreement |
The compliance row does the most work. OpenAI signs a ChatGPT BAA only for sales-managed Enterprise and Edu accounts, and its help center says outright that no BAA is offered for Business. For any organization touching patient data, that line usually settles the comparison before seat math enters it.
For everyone else, the honest crossover test is operational. If IT provisions accounts by hand and 2 to 149 people need seats, Business covers it, and the full consumer-and-team lineup is broken down in CloudZero’s ChatGPT pricing guide. When SCIM automated user provisioning, residency, or workforce-wide deployment enters the conversation, you’re an Enterprise buyer and the price becomes a negotiation.
One scoping distinction saves confusion later: none of this includes the API. Developer access is metered separately, with its own rates, covered in the OpenAI API pricing guide. A company running both pays both bills.
How do you forecast ChatGPT Enterprise spend before signing?
Forecasting ChatGPT Enterprise spend starts with the contract, not the rate card, because the public card is a list-price map rather than a copy of your invoice. Before modeling anything, record seven terms from the draft agreement: the seat or platform fee, committed spend, included usage, the discount schedule, overage rates, whether regional processing applies, and renewal mechanics.
Then model usage across a wide range, not a single average. Token-metered seats don’t behave like licenses. The spread between a light chat user and a Codex-heavy power user isn’t a rounding error, because agentic tasks chain model calls and fast mode multiplies each one. That’s why the $60 seat average hides more than it reveals.
A forecasting sequence that holds up in negotiation:
- Segment your workforce by intended usage: chat-only, document tools, and agentic Codex or Work users.
- Price a realistic month for each segment against the rate card, multipliers included.
- Compare that total to the included usage in the draft contract, since the gap between them is your true overage exposure.
- Confirm in writing which rate card governs the workspace, token or credit, because a forecast built on the wrong one fails without warning.
- Set workspace budgets and per-user limits in OpenAI’s admin console on day one, before usage patterns set.
Model churn belongs on the checklist too. OpenAI retired GPT-5.4 and GPT-5.4 mini in Codex for users signed in with ChatGPT on August 31, 2026, and the model lineup has rotated all year. Your negotiated rates stay fixed while the models underneath them change, which shifts what a “typical” task costs between the day you sign and the day you renew.
Anthropic and Google run the same playbook at the enterprise tier: sales-gated pricing, published usage rates underneath. The per-plan details live in CloudZero’s Claude pricing and Gemini pricing guides, with the cross-provider view in the LLM API pricing comparison. Since none of the three lists an enterprise price, the negotiating edge comes from showing up with your own usage data.
How do finance teams keep enterprise AI spend accountable?
Once an Enterprise workspace bills by the token, it behaves like cloud infrastructure: variable, usage-driven, and sensitive to configuration choices nobody runs past finance. Fast mode, residency, model selection, and agent concurrency all move the invoice, and they’re toggled by individual employees, thousands of times a day.
Seat-level invoices can’t answer the question that follows. “We have entered the ‘prove it’ phase,” PwC wrote in May 2024, when its US and UK firms announced a rollout to more than 100,000 employees across the US and UK. Two years later, proving it means knowing what each user’s tokens cost and what they produced.
That’s now directly measurable. CloudZero’s OpenAI integration connects at both layers: the platform connection brings in API costs by project, model, and operation, and the Enterprise connection pulls per-user Codex costs straight from your ChatGPT workspace, with Chat and Work usage flowing in through OpenAI’s Compliance API at hourly granularity and a full token breakdown, with no double counting between the two.
From there, enterprise ChatGPT spend gets the same treatment as the rest of the AI portfolio.
The AI Hub normalizes it alongside Azure OpenAI, Anthropic, and cloud spend, allocated to teams, features, and customers without tagging.
Budgets and forecasting turn negotiated commitments into forward plans, and anomaly detection flags the fast-mode surge weeks before the invoice lands.
Leading global teams like Skyscanner, Upstart, and Remitly use CloudZero to walk into renewals with numbers instead of impressions. There’s a real difference between telling a board “we spend a lot on OpenAI” and showing what Codex costs per merged pull request and what that work returned. The first defends a budget. The second grows one.
Schedule a demo to see per-user ChatGPT Enterprise costs land next to your API and cloud spend, get a free cloud cost assessment, or take a self-guided tour to explore AI allocation in action.
Frequently asked questions about ChatGPT Enterprise pricing
