Learn
Codex Credits vs API Billing: ChatGPT Work and API Keys
Codex and ChatGPT Work share plan-side usage and credits. API-key Codex and direct OpenAI API calls remain a separate Platform budget.
Clarify the spend threshold before you commit. Use this page when the core product is familiar and the real question is whether to stay free, upgrade, or switch pricing tracks.
Editorial guide
Guide
Start with the spend threshold and the conditions that change the pricing decision.
Short answer: ChatGPT Work and Codex share plan-side agentic usage and eligible credits. Those credits are not API credits. API-key Codex, CI jobs, apps, and backend systems use a separate OpenAI Platform ledger and standard API pricing.
Start with Codex Pricing when you need the current plan ladder, credits, and task-limit details. Use Is Codex Free? for the free-access boundary, Codex ChatGPT vs API Access for sign-in routes, Codex CLI Cost for terminal workflows, and ChatGPT Subscription vs OpenAI API Pricing when the broader question is whether subscriptions cover developer API usage.
The four budget questions
Codex has several official access routes, so the cleanest way to budget it is to separate the question before looking at prices. Plan packaging asks what the user bought. Task limits ask how much Codex work the plan includes before the user sees a limit, credit option, reset, upgrade path, or workspace control. Credits ask how additional eligible Codex usage is drawn down after included usage. API billing asks what a Platform account owes for token-metered usage.
Those questions can involve related models, but they are not the same ledger. OpenAI describes Codex as included across ChatGPT plans, with usage limits and credit options that vary by plan. OpenAI also says API-key usage is billed through the OpenAI Platform account at standard API rates and that ChatGPT and API platform billing are managed separately.
Budget route table
Budget route | What it covers | How to think about cost | Do not confuse it with |
|---|---|---|---|
ChatGPT plan access | Free, Go, Plus, Pro, Business, Edu, or Enterprise access to Codex through supported ChatGPT-owned surfaces | A subscription or workspace entitlement with included usage, plan-specific limits, and possible upgrade or credit options | OpenAI API usage from an API key, app, backend service, or Platform project |
Codex CLI, desktop, and workspace access | Codex on surfaces such as web, the CLI, IDE extension, app, and managed Business or Enterprise workspaces | The same surface can fall under ChatGPT sign-in, Business workspace controls, Enterprise access tokens, or API-key authentication depending on how it is signed in | A standalone CLI-only price; the installed client is not the billing owner by itself |
API key usage in Codex | Local Codex workflows in the CLI, SDK, or IDE extension authenticated with an OpenAI API key | Standard API pricing applies, model availability follows the API key, and cloud or workspace-only Codex features may be unavailable | Included ChatGPT plan credits, Business workspace credits, or ChatGPT cloud-task access |
Developer API billing | Your own software calling OpenAI APIs directly for product features, automation, internal tools, or CI systems | Forecast by model, input tokens, cached input, output tokens, tools, retries, service tier, and volume in the Platform account | ChatGPT subscription capacity, Codex task windows, or human-user message limits |
Credits are an extension mechanism
For Work and Codex, credits extend eligible usage after the plan's included allowance; they are not a universal OpenAI wallet. Plus and Pro users can add credits for supported agentic features. Free and Go users are prompted to upgrade instead of buying Codex credits.
Business, Edu, and Enterprise add a workspace layer for credits, controls, audit visibility, SSO, and procurement. Standard Business seats include ChatGPT and Codex. Codex-only seats should appear in a budget only for workspaces with a seat before June 24, 2026 or a pending invite as of June 24.
The current Codex rate card is token-based and says GPT-5.6 usage averages 5–40 credits per message. Actual use depends on input, cached input, output, reasoning, and tools. This still belongs to the shared ChatGPT agentic budget and does not turn credits into API billing coverage.
Task limits are not API invoices
Task limits are product limits. They control how much Codex work a user can do in a period, across surfaces such as local messages, cloud tasks, and code reviews depending on the plan and model. Reaching a limit can lead to waiting, switching models, adding credits where eligible, using a reset, upgrading, or asking a workspace admin to adjust spend controls.
API billing is different. When Codex signs in with an API key, OpenAI says usage is billed through the Platform account at standard API rates instead of included ChatGPT plan credits. When your own product calls OpenAI APIs directly, the bill follows the API pricing page, not the Codex task-limit table. Playground usage is also API usage, so developer experiments can belong to the Platform budget even when they feel exploratory.
This distinction matters most for automation. A human asking Codex to help with a repository through ChatGPT sign-in is usually a plan, limit, and credit decision. A CI job, code generation service, background agent, or internal tool using an API key is a developer API budget decision. The first is constrained by subscription or workspace packaging; the second is constrained by model rates, tokens, projects, budgets, and usage monitoring.
Final buying check
Before budgeting, write the workflow in one sentence. If a person uses ChatGPT Work or Codex through ChatGPT desktop, CLI, IDE, web, or a workspace, start from plan access and shared usage. If software uses an API key or calls OpenAI endpoints, start from API billing.
For most mixed teams, the right answer is two budget lines rather than one blended number: Codex subscription or workspace access for people, and OpenAI API usage for systems. That prevents the two common mistakes: assuming a ChatGPT upgrade will cover API calls, or estimating a developer automation project from human-facing Codex task limits.
Evidence boundary
Official sources
Editorial guidance grounded in official product sources.
FAQ
Common questions
Are Codex credits the same as OpenAI API billing?
No. ChatGPT Work and Codex can share eligible plan-side credits; those credits do not pay API invoices. API-key and direct API usage are charged through the OpenAI Platform.
Does a ChatGPT plan cover Codex API-key usage?
No. OpenAI says Codex can use ChatGPT sign-in for subscription access or an API key for usage-based access. API-key usage uses standard API pricing instead of included ChatGPT plan credits.
Are Codex task limits the same as token prices?
No. Task limits are product limits for Codex usage under a plan or workspace. Token prices are the developer API meter used by the Platform account for API calls and API-key workflows.
How should a Business team budget Codex and ChatGPT Work?
Use standard Business seats plus shared workspace credits for Work and Codex, and keep API usage as a separate budget. Only workspaces with a Codex seat before June 24, 2026 or a pending invite as of June 24 should include legacy Codex-only seats.
When does Codex move into the API budget?
It moves into the API budget when the workflow signs in with an API key or your own software calls OpenAI APIs directly. Common examples include CI jobs, backend agents, internal automation, SDK workflows, and product features.
What is the safest budgeting shortcut?
Use one line for Codex subscription or workspace access for people, and a separate line for OpenAI API usage for systems. Then check the Codex pricing guide and API pricing page for the current limits and rates before paying.
Next steps
Take the next buying step
Use these next pages to confirm the plan, tool, or alternate route that fits once the spend boundary is clear.