Custom GPT Cost: Build Scope, Integrations and Maintenance

Custom GPT pricing has two parts. OpenAI charges nothing per GPT: building and publishing GPTs is included in ChatGPT Business (formerly ChatGPT Team), Enterprise and Edu seats and is no longer available on Free, Go, Plus or Pro, so the license cost is the seat price of the plan your people use, $20 a month per Standard seat billed annually or $25 monthly on Business as of October 2026. The project cost is what you pay someone to scope, build, test and maintain the GPT, and that depends on whether it needs prepared knowledge, whether it connects to business systems, and who keeps it working afterward. One more fact up front: OpenAI has scheduled custom GPTs to retire on December 11, 2026 in favor of plugins and workspace agents, so a GPT quoted today should be scoped as its successor from the start. The cost drivers do not change; the container does.

NetSys quotes each build as a fixed project after a scoping call and publishes no price list, so this guide gives you the structure of a quote: what the license covers, what moves the project price, what continues after launch, and what to bring.

What do you pay OpenAI for a custom GPT?

Nothing beyond the plan. OpenAI's article on creating and editing GPTs says new GPT creation and publishing are not available on personal accounts, including Free, Go, Plus and Pro, and that users in Business, Enterprise and Edu workspaces can create, edit and publish GPTs when workspace settings and permissions allow. Prices as of October 2026, from OpenAI's business pricing page and its ChatGPT Business overview:

Core use of a GPT needs no credits; credits pay for advanced features such as deep research (50 credits a task as of October 2026) and ChatGPT Work usage, and owners can cap them per seat type. If the plan decision is open, our guide to ChatGPT Business pricing and setup cost covers seats, billing and the Enterprise line.

What drives the cost of building a custom GPT?

Three tiers of scope, and the price rises with each because the work moves from writing to data preparation to software:

The first tier is the smallest build and one most companies can do themselves after training. The second costs more than people expect because the GPT is only as good as the documents, and most companies have three versions of every policy. The third is a software project with a maintenance line, and the one most affected by the retirement date, because GPT actions do not transfer to plugins.

Does a plugin or workspace agent cost more than a custom GPT?

Not in license terms; the seats are the same. What changes is who can build, where the project effort goes and whether a usage line appears. OpenAI's custom GPT retirement and migration FAQ sets the dates (December 11, 2026; February 11, 2027 for Enterprise workspaces with an approved deferral; new creation on Enterprise planned to end October 26, 2026) and the migration path (instructions become a skill, knowledge files become reference files, actions are rebuilt, the migrated plugin starts private); its articles on custom MCP servers and workspace agents supply the rest, as of October 2026:

So a plugin with reference files quotes like a knowledge-prepared GPT, a workspace agent with a custom MCP server like an integration-connected GPT plus a usage budget, and nothing has to be rebuilt in December.

What does knowledge preparation involve?

Four tasks, each billable time: an inventory of which documents answer the questions in scope and which version is current; cleaning out drafts, superseded policies and anything a reader should not see, since anyone who can chat with the GPT can ask what the files say; converting the set into supported formats and splitting long manuals into parts the model retrieves well; and written confirmation that every file is cleared for everyone the GPT is shared with. OpenAI notes that GPT builders cannot view users' conversations with their GPTs, so the files are your only control over what it can reveal. Knowledge files are retained until the GPT is deleted and are copied into the plugin's reference files on migration.

What do integrations add to the price?

An integration adds a second system to build and run. Whether it is a GPT Action today or a custom MCP server in a plugin, the cost lines are the same: an endpoint you own in front of the business system, a least-privilege credential, OAuth so each user is identified, confirmation on writes, tests against hostile input, hosting, and someone reading the logs. Our guide to custom GPT Actions and safe business integrations walks through the build. If the job must run without a person in the chat, the build becomes a workspace agent or an application on the API, where usage is billed per token on top of the seats: as of October 2026 OpenAI's API pricing page lists GPT-6.1 Sol at $2.00 per million input tokens and $10.00 per million output tokens, and GPT-6 Luna at $0.10 and $0.50, and a Business subscription does not include API usage.

Which costs continue after launch?

• Evaluation. Real questions with expected answers, rerun when the model, the documents or the integration change; OpenAI updates models on its own schedule.
• Employee handoff. A short session for the users and a named owner who collects change requests; our AI training for employees covers this when a company wants it run formally.
• Changes and migration. New products, forms and policies, each a small change with a test run, and the move to a plugin before December 11, 2026.
• Monitoring for integrations. Logs, credential rotation, cost against budget and a monthly review, which our managed AI services agreement covers month to month.

Custom GPT cost examples

Three illustrative situations, written for this page rather than taken from clients, with license figures at OpenAI's list prices as of October 2026 and project lines that describe scope rather than a figure.

• An eight-person firm wants a handbook and benefits assistant. License: eight Standard seats on ChatGPT Business billed annually, 8 x $20 = $160 a month. Project: instructions plus knowledge preparation, a test set and one training session. Ongoing: a quarterly document refresh. Successor: a plugin with reference files.
• A 25-person distributor wants order status in ChatGPT. License: 25 x $20 = $500 a month annually. Project: a read-only gateway in front of the order system, OAuth, test cases and a pilot with the sales desk. Ongoing: monthly monitoring. Successor: a plugin with a custom MCP server, or a workspace agent if it should also run on a schedule.
• A 60-person company needs audit logs of every conversation. License: ChatGPT Enterprise, quoted by OpenAI, because the Compliance Platform is Enterprise and Edu only. Project: as above plus role-based publishing and the action domain allowlist. Ongoing: a monthly review that includes the compliance logs.

When is an application on the API cheaper than a GPT?

When the users are outside your workspace, when the job runs on a schedule or at volume, or when the output must land in your own systems. A GPT costs a seat per person who uses it; an API application costs tokens per run plus hosting, with no seat for someone who never opens ChatGPT. The two also differ in ownership and support, so we scope them as separate offers.

Quote checklist: what to bring for a fixed price

1. The one job the GPT should do, with five real examples of questions and acceptable answers.
2. Who will use it (headcount and roles) and which ChatGPT plan you have or expect to buy.
3. The documents involved, who owns each and when it was last updated.
4. The systems it must read or write, whether each has an API and who administers it, plus any data restrictions (client, health or financial information).
5. Who inside the business will own it after launch, and how you will judge success three months in.
6. Your position on the retirement: build as a GPT and migrate, or build as a plugin or workspace agent now.

How NetSys prices custom GPT work

Our custom GPT development service starts with a scoping call that produces the one-page specification above, then a fixed project price for the build and an optional month-to-month line for monitoring and changes. Integrations are built under our OpenAI and ChatGPT integration service with the gateway, credentials and tests described here, and every GPT we quote now includes the plan for its successor.

Request a scoped GPT quote and bring the checklist.

Frequently asked questions

How much does a custom GPT cost?

From OpenAI, nothing beyond the ChatGPT plan: $20 or $25 a month per Standard seat on Business as of October 2026, or an Enterprise quote. From a builder, a fixed project whose price follows the scope: instructions only, prepared knowledge, or a connected integration, plus optional monthly upkeep. NetSys quotes after a scoping call and publishes no price list.

Is there a fee per custom GPT or per conversation?

No. OpenAI prices seats, and credits only for advanced features such as deep research. An application built on the API is different: it is billed per token used.

Do we need ChatGPT Business to build a custom GPT?

Yes, a workspace plan: Business, Enterprise or Edu. OpenAI says new GPT creation and publishing are not available on Free, Go, Plus or Pro accounts. Business needs at least two paid seats.

Should we still build a custom GPT in late 2026?

Build the job, not the container. OpenAI retires custom GPTs on December 11, 2026 and recommends plugins; workspace agents cover jobs that run on their own. If a GPT is the only thing your workspace supports today, build it with the migration planned, and never build an action you are not prepared to rebuild as an MCP server.