TL;DR: ChatGPT Enterprise has no public list price, but buyers report about $60 per seat per month, a 150-seat minimum and annual contracts, which puts entry spend near $108,000 a year. On top of seats, token-based usage is metered and billed separately, API usage is a second bill, and integrations, rollout and governance add internal cost, so real TCO can run 50% or more above the seat quote. Usage and integration lines are the most variable: model them with a low, expected, and high case before contract approval.
ChatGPT Enterprise usually costs around $60 per seat per month, but the seat price is only the starting point. Once you add the seat minimum, annual commitments, metered usage, integrations and rollout work, the real total cost of ownership (TCO) can look very different from the quote.
This guide breaks down what OpenAI actually charges, where the hidden costs sit, and how to model the full cost before the contract reaches approval.
What ChatGPT Enterprise costs per seat
OpenAI does not publish a list price for Enterprise. It is quote-based, and buyers report an average of about $60 per user per month, within a $45 to $75 range, according to CloudZero. OpenAI has not confirmed these figures.
Volume moves the price. One licensing advisory's mid-2026 benchmarks, cited by CloudZero, put seats at:
Seats | Reported price per seat per month |
|---|---|
150 to 999 | $55 to $60 |
1,000 to 4,999 | $48 to $55 |
5,000+ | $40 to $48 |
The commercial terms matter as much as the rate:
Seat minimum: commonly reported at 150 seats, though not published by OpenAI.
Contract length: annual, with no month-to-month option.
Entry floor: about $108,000 per year (150 seats x $60 x 12). QAnswer similarly puts first-year entry spend at $100k+.
Discounts: up to 75% off for nonprofits; volume and multi-year discounts are negotiated.
ChatGPT Business vs Enterprise
For teams under the Enterprise minimum, ChatGPT Business is the self-serve alternative at a fraction of the commitment. The trade-off is fewer contractual and compliance options.
ChatGPT Enterprise | ||
|---|---|---|
Standard seat | $20/month billed annually, $25 billed monthly | ~$60/month reported, quote-based |
Premium seat | $100/month billed annually, $125 billed monthly | Negotiated |
Seat minimum | 2 | ~150 (reported) |
Buying | Self-serve | Sales-led annual contract |
BAA (HIPAA) | Not offered | Available on sales-managed Enterprise and Edu |
API usage | Billed separately | Billed separately |
Business includes GPTs, Projects, Apps, Company Knowledge, ChatGPT Agent and Deep Research, plus admin and spend controls, and OpenAI does not train on workspace data. Invoice billing, purchase orders, net terms, Zero Data Retention and BAAs require a contracted offering. OpenAI signs BAAs only for sales-managed Enterprise and Edu accounts.
For context, QAnswer reports that Microsoft Copilot, Google Gemini Enterprise and Claude Team price closer to $30 per seat, roughly half to a third of a typical ChatGPT Enterprise seat.
The hidden costs beyond the seat
The biggest TCO risk is metered usage: on token-based agreements, seat fees and usage are billed separately. Here is where the extra spend comes from.
1. Token-based usage
On eligible Enterprise agreements, usage across Chat, Work and Codex is metered in tokens and charged in US dollars, separate from seat fees (OpenAI). Seat fees are based on assigned seats at the start of each month, even if those users are inactive, and usage is invoiced monthly in arrears. A spending commitment is not a prepaid balance, and any shortfall is billed under your agreement.
Sample rates from OpenAI's Enterprise rate card (USD per 1M tokens):
Model | Input | Cached input | Output |
|---|---|---|---|
GPT-6 Pro (GPT-6 Astra) | $10.00 | $1.00 | $50.00 |
GPT-5.6 Sol | $4.00 | $0.40 | $20.00 |
GPT-6 (GPT-6 Sol) | $2.00 | $0.20 | $10.00 |
GPT-5.6 Luna | $0.20 | $0.02 | $1.20 |
2. Feature charges and multipliers
Some features carry their own fees on top of model tokens:
Web search: $10 per 1,000 runs, plus content tokens.
Voice: $0.05 per minute, plus model tokens.
Long context (over 272K input tokens): 2x input, 1.5x output.
Fast mode: 2x; Ultrafast on GPT-6 Astra: 6x.
Data residency (regional processing): 1.1x on standard rates.
3. Billing model changes
Some promotional rates, such as GPT-5.6 Sol for Work and Codex, run only through at least November 21, 2026. Budget for rates to change at renewal.
4. API usage
API usage is not covered by ChatGPT seats and is invoiced separately. Any internal app or agent built on OpenAI models adds a second bill.
5. Integrations and engineering
Custom connectors to internal systems usually mean extra engineering work on your side. Document-level permissions tied to source systems also have to be built around the product, not inside it.
6. Rollout, training and governance
None of the sources cited in this guide publish standard figures for implementation, training, or change management. Estimate these internally: admin time, enablement sessions, security review and legal review of data-residency terms. QAnswer notes that US-hosted contracts can be a hard blocker for regulated sectors.
Worked example: First-year TCO for 300 seats
In this illustrative model, a 300-seat company pays about $207,000 for seats but roughly $310,000 all-in, or about $86 per seat per month. Only the seat price is sourced; the other lines are assumptions to replace with your own numbers.
Cost line | Basis | Year 1 |
|---|---|---|
Seats | 300 x $57.50 midpoint of the $55 to $60 benchmark x 12 | $207,000 |
Metered usage above included amounts | Assumption: $10 per user per month | $36,000 |
API for internal tools | Assumption: $1,000 per month | $12,000 |
Integration engineering | Assumption: 0.25 FTE at $160,000 | $40,000 |
Rollout, training and governance | Assumption: one-off enablement and review | $15,000 |
Total | $310,000 |
The gap between seat price and true cost is about 50% in this scenario. Usage and integration lines are the most variable, so model them with a low, expected and high case.
How to reduce TCO and negotiate
The biggest savings come from right-sizing seats and capping usage before the contract starts.
Start on Business if you are under 150 users. Business starts at $20 per seat with a two-seat minimum.
Negotiate on volume and term. Per-seat benchmarks drop from $55 to $60 at 150 seats to $40 to $48 above 5,000, and multi-year discounts are negotiable.
Remove inactive seats monthly. Seat fees are based on assigned seats at the start of each month, even for inactive users; removed seats reduce the next month's baseline.
Set budgets and limits. Admins can set a monthly workspace budget and user or group limits.
Steer usage to cheaper models. GPT-5.6 Luna costs $0.20 per 1M input tokens versus $10 for GPT-6 Astra.
Limit multipliers. Restrict fast mode and long-context use to teams that need them.
Ask for nonprofit pricing if eligible. Discounts reach up to 75%.
Cutting integration costs with Composio
Composio lets ChatGPT work with 1,500+ apps through one connection, which can shrink the integration line of your TCO. Instead of funding and maintaining a custom connector for every tool, IT connects ChatGPT once and gets access to a large catalog of SaaS toolkits.
Composio supports apps such as Gmail, Google Calendar, Google Drive, Google Docs, Slack, Notion, GitHub, Salesforce and HubSpot. That means ChatGPT can read a thread, update a CRM record or file a GitHub issue from the same chat.
You can install the Composio plugin from ChatGPT's plugin marketplace, or add Composio as a custom MCP server using its Streamable HTTP endpoint. A single MCP endpoint gives access to many apps, and Composio loads only the tools relevant to the task instead of every tool at once.
Why it lowers TCO:
Composio provides one governed path to act across 1,500+ business systems. It offers over 50,000 agent-ready tools, 1 million+ connected accounts and 1B+ tool calls processed, so IT governs one integration instead of many.
Every action is traced (who, what, why, result), which is what lets IT sign off on agent deployments. Composio is SOC 2 Type II certified, encrypts tokens at rest with AES-256 and in transit, and lets admins limit allowed scopes and actions.
Less connector engineering. Managed auth also handles OAuth, API keys, token refresh and scopes, so there's no auth layer for your team to build or maintain, and automations don't stall on expired credentials. User reviews on SoftwareAdvice and GetApp report Gmail and Google Drive integrations completed in under 30 minutes.
Its MCP Gateway adds tool-level controls and an audit trail for every call. If integrations are a large line in your model, as in the worked example above, pricing Composio against in-house connector work is a quick way to test how much of that cost you can remove.
ChatGPT Enterprise starts near $108,000 a year at the reported 150-seat minimum, but usage, API, integrations and rollout can add 50% or more on top. Treat the seat quote as one line in a wider model: track metered usage from day one, set budgets, and cut integration work with tools like Composio so more of your spend goes to value, not plumbing. Book a call to walk through your security requirements before the next enterprise review.
FAQs
How much does ChatGPT Enterprise cost?
OpenAI doesn't publish a price. Buyers report about $60 per user per month, within a $45 to $75 range.
What is the minimum number of seats?
Around 150 seats is commonly reported, though OpenAI doesn't publish it. ChatGPT Business needs only 2 seats.
Can I pay monthly?
Enterprise is an annual contract with no month-to-month option. On token-based agreements, seat fees and usage are invoiced monthly in arrears.
Is API usage included?
No. API usage is invoiced separately from ChatGPT seats and usage.
Do I pay for inactive users?
Yes. Seat fees are based on seats assigned at the start of each billing month, even if those users are inactive.
How do I connect ChatGPT to my other apps?
ChatGPT supports MCP servers and plugins. Composio provides one MCP endpoint that connects ChatGPT to 1,500+ app toolkits, with managed authentication.
Key terms glossary
API: Programmatic access to OpenAI models for building your own apps; billed separately from ChatGPT seats.
BAA: Business Associate Agreement, a contract required to handle protected health information under HIPAA.
Cached input: Input tokens the model has recently processed, charged at a lower rate.
Codex: OpenAI's coding agent, available in ChatGPT and metered by tokens on Enterprise.
Data residency: Processing and storing data in a chosen region; billed at a 1.1x multiplier on token rates.
MCP: Model Context Protocol, a shared specification for connecting AI models to external tools and data.
Zero Data Retention: An option where OpenAI doesn't store prompts and outputs; requires a contracted plan.
Get started
