What Are Custom GPTs? Who Can Build One and What It Costs

A custom GPT is a version of ChatGPT set up for one specific job. It combines saved instructions, reference files and optional tools for that job.

Personal accounts cannot create new GPTs, including Free, Go, Plus and Pro. Members of Business, Enterprise and Edu workspaces may create one when allowed. A workspace is the shared ChatGPT area for a company or school. Business Standard costs $25 per person monthly on monthly billing, with two paid people required. Last checked: US prices as listed on 26 October 2026.

OpenAI plans to retire custom GPTs on 11 December 2026. It plans to help creators move their work to plugins, which bundle instructions and connected tools for a task. Existing GPTs remain usable for now. You can use one without buying Business. If you made one, prepare its replacement before retirement.

Why do older guides still say Plus can make a GPT?

Many guides in the search results still describe a path through the GPT builder on Plus. OpenAI's current help page says new creation and publishing are unavailable on personal Free, Go, Plus and Pro accounts. That is the rule to use when deciding whether to pay for a plan today.

OpenAI's general pricing page still shows “custom GPTs” under Plus. The specific help page gives the narrower rule. Using an existing GPT and making a new GPT are different permissions. A Free user can discover and use GPTs, subject to the Free plan's limits. Paying for Plus alone does not unlock new GPT creation on a personal account.

OpenAI is also changing the product itself. Its retirement FAQ says the transition affects every ChatGPT plan, although migration and plugin access can vary by account or workspace. When a guide tells you to upgrade to Plus to build a GPT, check whether it predates these changes. Spend money on the plan you actually need, not on an old screenshot of the Create button.

What does a custom GPT actually do?

Think of it as a repeatable brief for ChatGPT. The instructions describe the task and how the assistant should respond. Knowledge files give it material to consult. Conversation starters show a new user what to ask. You can also enable capabilities such as web search or image generation. The result still lives inside ChatGPT.

A small firm might have a GPT that turns an approved product sheet into a first draft of a customer reply. The owner writes the standing instructions once, adds the current sheet and asks for a draft when a question comes in. Picture a school department using its style guide to draft letters in a consistent format. Those are examples of possible setups, not measured outcomes.

OpenAI lets a creator attach up to 20 knowledge files, with each file up to 512 MB. The distinction between instructions and knowledge matters. Instructions tell the GPT how to behave. Knowledge supplies information it should consult.

A GPT can connect to outside services through apps or through custom actions. An app connects a service you use, such as a shared drive. An action calls an outside API, a connection a developer may set up. OpenAI says one GPT can use either apps or actions, but not both at once. Check the service's own access and permissions before using it.

Who can create a new one now?

OpenAI separates personal accounts from managed workspaces. Free, Go, Plus and Pro are names of personal plans. People on those plans cannot create or publish a new GPT. People in Business, Enterprise or Edu workspaces can create, edit and publish when their workspace settings and role allow it. A workspace owner or administrator may decide which members have that permission.

The builder is on the web. Mobile apps can use GPTs, but do not offer GPT creation. An eligible member opens the GPTs area, chooses Create, then either describes the intended job to the builder or sets its fields directly. OpenAI's guide recommends testing in Preview, the builder's test area, before saving. Treat that test as a chance to catch bad instructions and missing source material, not as proof the GPT will always answer correctly.

Enterprise workspaces have an extra clock to watch. OpenAI's retirement FAQ lists 26 October 2026 as the planned end of new custom GPT creation for affected Enterprise workspaces. Anyone in a managed workspace should check the notice for that workspace before planning a new build.

How much does creating a GPT cost?

ChatGPT Business Standard costs $25 per user per month with monthly billing, or $20 per user per month when billed annually. A Business workspace requires at least two paying members. That means paying for two Standard users at the listed monthly price. Last checked: US list prices as listed on 26 October 2026.

Business also offers Premium seats at a higher price, but a Premium seat is not required merely to use the GPT builder. Enterprise and Edu access follows their workspace arrangements and permissions. The Business FAQ says its US prices can vary by country and currency, so the amount in your checkout may differ.

Plus is listed at $20 per month on OpenAI's individual pricing page. That price buys Plus features, but the current GPT creation guide excludes personal Plus accounts from making new GPTs. In our view, an older tutorial's Create button is a poor reason to pay for Plus.

What happens to GPTs already in use?

OpenAI says existing GPTs remain usable until their retirement date, subject to access and workspace permissions. The planned standard retirement is 11 December 2026. Affected Enterprise workspaces with an approved extension have a later date of 11 February 2027.

Using someone else's GPT calls for a simple check: who created it, and have they named a replacement? OpenAI says the account or workspace where a GPT was created controls its timeline, even when you open it from another account. Access to the original does not automatically grant access to the new plugin.

If you created the GPT, list the instructions, files and connected services it relies on. Keep a few familiar prompts and their useful answers so you can compare the replacement later. OpenAI says the migration copies instructions and knowledge files into a plugin, but the selected model does not carry over. Custom actions do not transfer automatically. A workflow that depends on one of those actions needs separate work before it can perform the same task.

Migration has another easy-to-miss consequence. The replacement plugin starts private. It is not installed for your users, and it does not inherit the GPT's sharing settings. After migrating, test the plugin with a person who actually uses the workflow.

Why is OpenAI replacing custom GPTs?

OpenAI's stated aim is to bring reusable instructions and connected apps together in plugins. An app connects another service; the saved instructions tell ChatGPT how to handle the job. The plugin can bring those pieces together beside the tools and information the task needs. OpenAI recommends moving GPT workflows to that format as it retires GPTs.

A plugin can package a reusable skill, connected apps or both. A skill is the saved set of instructions for a job. That packaging does not mean a migrated GPT will behave exactly as before. OpenAI asks creators to compare familiar prompts and test difficult cases after migration. The practical test is whether the new setup still produces useful work for its regular users.

For an owner who only needs a repeatable writing brief, the important part may be the instructions and reference files. For a team whose GPT reaches into another system, the app or custom action deserves extra attention. A polished answer is not enough if the task depended on a connection that no longer runs.

What should you do today?

If you only use GPTs made by others, keep using the ones available to you and watch for notices on the GPTs that matter. Free users can discover and use GPTs. You do not need to buy Business just to ask a GPT a question.

If you want to build something new, check whether your Business, Enterprise or Edu workspace still allows it. Remember that the product is scheduled for retirement. A new GPT may be useful as a short bridge for an existing team, but plan where the task will live afterward. For a new long-term workflow, examine the plugin path first.

If you maintain an existing GPT, make a small inventory: its purpose, owner, users, instructions, knowledge files and any outside action. Then test a migrated version when that option appears in your account. Keep the replacement private while you compare real tasks.

Frequently asked questions

Can I use a custom GPT on the Free plan?

Free users can discover and use GPTs. A personal Free account cannot create or publish a new GPT.

Can I make a new GPT with ChatGPT Plus?

OpenAI's current GPT creation guide says personal Plus accounts cannot create or publish new GPTs. The same restriction applies to personal Free, Go and Pro accounts.

What is the lowest listed Business cost for a new GPT?

Business requires at least two paid seats. Each Standard seat costs $25 per month on monthly billing at the listed US price. OpenAI says prices can vary by country and currency.

Are custom GPTs gone already?

Existing GPTs remain usable until the retirement date that applies to their account or workspace. OpenAI lists 11 December 2026 as the standard planned date and says an approved Enterprise extension can last until 11 February 2027.

Will moving a GPT to a plugin keep its actions and sharing?

OpenAI says custom actions do not transfer automatically. The replacement plugin starts private and does not carry over the GPT's sharing settings, so test its connections and grant access separately.

Sources

OpenAI: GPTs in ChatGPT; OpenAI: Creating and editing GPTs; OpenAI: Free Tier FAQ; OpenAI: retirement and migration FAQ; OpenAI: Business FAQ; OpenAI: plan prices; OpenAI: workspace access; OpenAI: plugins.