ChatGPT pricing in 2026: individual plans, Business seats, and API costs

ChatGPT pricing at a glance

Sources: OpenAI's plan comparison, Go guide, Plus guide, Pro tiers, and Business overview.

The Go help page directs readers to current checkout pricing. Do not assume a launch price in one market applies everywhere. Likewise, a monthly equivalent on an annual plan is not a month-to-month cancellation arrangement.

Individual plans: what changes as you upgrade

Free and Go

Free is a useful starting point for seeing whether ChatGPT helps with your actual tasks. The current comparison includes chat and selected tools with limits. A free account does not establish that every research, coding, image, or connected-app feature is available without restriction.

Go expands access to eligible features such as uploads, image creation, and data analysis. Some advanced capabilities still require Plus or another eligible plan. Check the feature you need rather than buying solely because a plan offers “more.”

This menu image comes from OpenAI's Go guide. Seeing a tool in an example interface is not a promise about its limit or availability in your account.

Plus

At $20/month, Plus offers broader access than Free, including eligible reasoning, voice, image, file-analysis, and research features. OpenAI notes that limits can vary with conditions. It is an individual subscription, not a substitute for centrally managing a company's seats and access.

Use a small set of representative tasks to judge the upgrade: an actual-length document, the kind of analysis you repeat, and the model or tool you expect to use. A feature list alone does not tell you whether your workload fits within the allowance.

Pro

Pro now has two price levels. The $100 and $200 tiers share core capabilities but offer different usage allowances. OpenAI also documents separate allowances for some models, including temporary unavailability after a limit is reached.

Choose based on observed usage, not an assumption that the most expensive subscription makes every task unlimited. Compare the extra allowance with how often you actually encounter a limit and whether another supported option meets the job.

Business pricing: seats, usage, and administration

ChatGPT Team was renamed ChatGPT Business. Current Business pricing distinguishes Standard from Premium seats. A workspace needs at least two seats in total, and can mix the two types.

Business includes centralized billing and administration. Standard and Premium provide access to ChatGPT, ChatGPT Work, and Codex; Premium adds higher usage. Workspace credits can extend usage beyond included limits. Review those terms when estimating the bill, rather than multiplying seat price and calling the result a complete forecast.

For example, two Standard seats at the published annual rate represent $40 per month equivalent, or $480 for the year before applicable tax. The monthly-billed equivalent is $50 each month. That is seat-cost arithmetic, not a quote for additional usage or services.

Enterprise is contract-priced. Ask for the controls, support, deployment terms, and usage arrangements your organization needs. An unofficial “around $60 per seat” estimate is not a sound basis for that decision.

The API is a separate cost

A ChatGPT subscription does not include API usage. The API pricing reference separates model input, output, applicable caching charges, processing options, and tool costs. Different modalities can have different units.

For a text application, build the estimate from measured requests: the selected model and processing mode, input and output token counts, cache behavior where applicable, and tool calls. Then add the application's other costs, such as retrieval, hosting, monitoring, and maintenance. Do not use a fixed words-to-tokens conversion as if every language and document format had the same ratio.

Usage-based billing is not inherently impossible to forecast. Measure a representative sample, include busy-period scenarios and retries, and compare the estimate with actual usage. A subscription and an API application should be compared for the same job, not merely by their headline prices.

Compare the cost of completed support work

Support work might involve retrieving a policy, looking up a record, drafting a reply, and handing an exception to a person. ChatGPT's current business offerings include connected apps and workspace workflows; it is inaccurate to claim that it can only copy text from a public-information chat window.

The useful comparison is your configured workflow. What reaches the agent? Which actions can run? What requires approval? How much review is still needed? Test those questions before promising savings or treating a model subscription as the entire operating cost.

The current eesel pricing documentation lists default pay-as-you-go pricing with no per-seat fee or monthly minimum: light tasks are free, regular tasks are $0.40, and heavy tasks are $4. A task covers the interaction rather than each reply. Annual commitments and Enterprise have separate terms; the latter adds a platform fee to usage.

That is not a claim that eesel is always cheaper. Compare the actual task mix, scope, and quality of the result. Tasks can be billed even when their outcome needs correction. Also distinguish dashboard questions from CLI chats: the CLI reference says inspection and setup commands are free, while chat is billed work.

Inspect the work behind a bill with eesel CLI

Suppose a support owner sees recurring work from an old hourly summary that the team no longer uses. A person at a terminal, a script, or a coding agent such as Claude Code, Codex, or Cursor can use eesel CLI to inspect the configuration and propose a change. The CLI operates the same teammate and workspace as the dashboard, and its JSON output can be reviewed without copying numbers from screenshots.

With Node.js 18.17 or newer, begin with read-only commands. Replace the example teammate name with the intended existing agent:

npx @eesel/cli whoami
npx @eesel/cli billing --agent "Support operations"
npx @eesel/cli automations list --agent "Support operations"
npx @eesel/cli activity --agent "Support operations"

Ask the coding agent to identify the recurring automation, its current state, and the activity that makes it worth reviewing. It should report uncertainty if the records do not establish which work caused a charge. It should not manufacture a per-run cost by dividing the whole invoice by a count of messages.

Preview a targeted pause

The owner first confirms that the selected schedule is obsolete and that pausing it will not remove a required handoff or report. Get its actual ID from the automation output and inspect the supported command:

npx @eesel/cli automations --help

For a scheduled automation, the documented editor supports automations update with --enabled false. Replace the placeholder below with the verified ID, then preview the change:

npx @eesel/cli automations update SCHEDULE_ID --enabled false --agent "Support operations" --dry-run

--dry-run prints the server call without sending the write. After the owner approves that exact target and consequence, repeat the command without --dry-run. Read back automations list --agent "Support operations" and check the affected workflow in the dashboard. This is a targeted pause, not deletion of every automation and not a change to billing terms.

Check the result, not just the receipt

Confirm that the intended automation is disabled while other required work remains enabled. Review later activity when the old schedule would have run, accounting for work already underway. If the owner wants to resume it, review the exact setting and approval again; do not assume a paused workflow should restart automatically.

If changing the scope of customer-facing work instead, run separate normal and excluded-case tests with approved costs and data. Disable consequential production actions during preliminary tests, then separately check the real channel with permission. A lower bill is not success if valid customer requests silently lose their response path.

Keep the budget connected to the workflow with eesel CLI

The screenshot shows an example workspace, not your account's current balance or limit. eesel's documentation describes a configurable monthly spend cap: new work pauses at the limit, while work already in progress finishes. Review your actual settings and the service impact of pausing work.

The CLI's billing command is read-only; changing payment or spending settings belongs in Billing & Usage. Use the CLI to gather evidence and review automation scope, and the dashboard to inspect the same setup. Try eesel to connect that review to the teammate doing the work.