What if you could use OpenAI Codex without buying a separate coding-agent subscription?
That is the first thing most people get wrong about OpenAI Codex pricing in 2026. Codex is no longer just an API model or a developer-only experiment. OpenAI now positions it as a coding agent that works across the Codex app, CLI, IDE extension, web, and ChatGPT-connected workflows.
The short answer as of June 9, 2026: Codex is included with eligible ChatGPT plans, including Free, Go, Plus, Pro, Business, Edu, and Enterprise. The catch is usage. Free and lower-tier plans get access, but heavier work, longer sessions, fast mode, agentic workspace features, and team usage can consume more Codex credits or require higher plan access.
If you are choosing between OpenAI and Anthropic for coding work, read this pricing guide first, then compare it with our Claude Code vs OpenAI Codex 2026 and OpenAI Codex App Guide 2026. If you are not a developer and want to know whether Codex is becoming a business productivity tool, read Codex for Knowledge Work 2026.
This guide covers the real buying decision: free vs paid Codex, ChatGPT plan access, Codex credits, API pricing, when Plus is enough, when Pro makes sense, when Business matters, and how the June 2026 Codex update changes the value calculation.
What Is the Quick Verdict on OpenAI Codex Pricing?
Most people should start with the Codex access already included in their ChatGPT plan before buying anything extra.
Here is the practical decision:
| User type | Best starting point | Why |
|---|---|---|
| Curious beginner | ChatGPT Free with Codex | Test whether agentic coding fits you |
| Solo developer | ChatGPT Plus ($20) | Better everyday limits and access |
| Serious builder | ChatGPT Pro $100 | 5x higher usage than Plus |
| Heavy builder | ChatGPT Pro $200 | 20x higher usage than Plus |
| Startup team | Business ChatGPT & Codex | Shared workspace, controls, Sites, plugins, and higher usage options |
| Development-focused team | Business Codex / Codex credits | Team-scale usage, review loops, and credit control |
| Enterprise team | Enterprise or Edu | Admin controls, compliance, and governed rollout |
| API builder | GPT-5.3-Codex API | Separate dollar-denominated API token pricing |
The important part: ChatGPT subscription access, Codex credits, and OpenAI API billing are related but not identical. If you use Codex through ChatGPT, the app, CLI, IDE, or Codex web, you are using plan-based access with credit/token metering. If you call a Codex model through the OpenAI API, you are using separate API billing.
Is OpenAI Codex Free in 2026?
Yes, Codex is available on ChatGPT Free, but “free” does not mean unlimited.
OpenAI’s current Codex with ChatGPT plan help page says Codex is included across Free, Go, Plus, Pro, Business, Edu, and Enterprise plans. It also says usage limits vary by plan. The ChatGPT pricing page describes Free as having limited Codex access, Plus as having expanded Codex usage, and Pro as having maximum Codex tasks.
That means the free tier is good for learning, small edits, experiments, and deciding whether you like the workflow. It is not a serious production plan.
Free Codex is not the best fit for:
- Large repository migrations
- Multiple long-running tasks
- Heavy daily coding
- Team workflows
- Production pull request review loops
- Running several agents in parallel
Use the free tier to answer one question: “Do I want a coding agent in my workflow?” If the answer is yes, upgrade based on how often Codex saves real work time. If you need to buy extra Codex credits, OpenAI’s credits help page says that is currently a Plus/Pro path; Free and Go users are pushed toward upgrading instead.
How Does Codex Work With ChatGPT Plans?
Codex uses your ChatGPT account for access. OpenAI’s help page lists several Codex clients you can launch with that account:
- Codex app
- Codex CLI
- Codex IDE extension
- Codex web
- ChatGPT-connected Codex workflows
This is why the pricing decision is different from older AI coding tools. You are not only paying for autocomplete. You are paying for a broader agent workspace connected to the OpenAI ecosystem.
The Codex app announcement is especially important because OpenAI describes the app as a command center for agents. It supports multiple agents in separate threads, worktrees for isolated changes, reviewable diffs, and continuity across surfaces.
That means the value of Codex is not just “Can it write code?” The better question is: “Can it move work forward while I supervise?”
What Changed After OpenAI’s June 2 Codex Update?
OpenAI’s June 2, 2026 Codex update matters because it changes what you are paying for.
Before this update, most users judged Codex like a coding assistant: “Can it edit my repository better than Claude Code, Cursor, or GitHub Copilot?” That is still a real question. But OpenAI is now positioning Codex as a broader agent workspace for developers and non-developers.
The update introduced or expanded three pricing-relevant pieces:
| New Codex capability | Who should care | Pricing implication |
|---|---|---|
| Role-specific plugins | Analysts, marketers, sales teams, finance, product, operations | Business and Enterprise plans become more relevant because workspace controls and app permissions matter |
| Sites in Codex | Teams that want dashboards, planners, review spaces, and internal tools | More reason to compare Business/Enterprise, not only Plus vs Pro |
| Annotations | People reviewing docs, dashboards, websites, slides, and app drafts | Codex becomes more useful after the first draft, which can justify higher usage if it saves review cycles |
OpenAI says the new role-specific plugins include 62 popular apps and 110 skills. It also says Sites are starting in preview for Business and Enterprise customers, letting teams create and share hosted interactive websites and apps inside a workspace. OpenAI’s separate knowledge-work report says Codex now has more than 5 million weekly active users, with knowledge workers representing about 20 percent of users and growing faster than developers.
The pricing takeaway is simple: individuals should still start with Free or Plus, but teams should now evaluate Codex as a workflow product, not only a coding model.
Which ChatGPT Plan Should You Use for Codex?
Here is the plain-English version.
| Plan | Codex access | Best for | Watch out for |
|---|---|---|---|
| Free | Limited Codex access | Testing, small tasks | You may hit limits quickly |
| Go | More access than Free where available | Lower-cost access markets | Availability varies by region |
| Plus ($20) | Expanded Codex usage | Solo developers and creators | Not ideal for constant heavy usage |
| Pro $100 | 5x higher usage than Plus | Real projects throughout the week | Still not for unattended abuse |
| Pro $200 | 20x higher usage than Plus | Heavy parallel workflows | Worth it only if usage is real |
| Business ChatGPT & Codex | Workspace plan with Codex | Teams and startups | Best when people actually share workflows |
| Business Codex / credit plans | Team-scale Codex usage | Development-focused teams | Costs vary by model, task size, and usage |
| Edu | Included for eligible institutions | Schools and research teams | Admin rollout matters |
| Enterprise | Included with enterprise controls | Larger companies | Requires governance and onboarding |
Start with Plus if you are an individual who codes or builds side projects regularly. Move to Pro $100 when you use Codex for real projects throughout the week. Use Pro $200 only when you run heavy workflows continuously or across parallel projects.
For teams, there are now two distinct questions. First, do you want ChatGPT plus Codex in one governed workspace? Second, do you need extra Codex credits because engineering work consumes the included allowance too quickly? The June 2026 Sites and plugin updates make that first question more important.
For many teams, workspace controls and repeatable workflows matter more than one person’s higher ceiling.
How Do Codex Credits Work Now?
This is the biggest pricing change to understand.
OpenAI’s Codex rate card says Codex pricing moved to a token-based credit model in April 2026 for Plus, Pro, Business, Enterprise, Edu, Health, Gov, and related plans. In plain English, Codex tasks are now easier to map to the same basic ingredients used in API pricing:
- Input tokens
- Cached input tokens
- Output tokens
- Model choice
- Fast mode or normal mode
- Code review or task type
The rate card gives credit rates per 1M tokens. For example, GPT-5.3-Codex uses 43.75 credits per 1M input tokens, 4.375 credits per 1M cached input tokens, and 350 credits per 1M output tokens. OpenAI also notes that code review uses GPT-5.3-Codex. The rate card also lists newer non-Codex and preview options, but you should budget around the live page because model availability and credit rates can change.
This does not mean every user needs to calculate tokens manually. Your included plan usage is used first, then eligible Plus and Pro users can add credits after hitting limits. It means the old question “How many messages do I get?” is less useful. The better question is: “How large and output-heavy are my Codex tasks?”
Light tasks consume less. Large repo tasks, output-heavy refactors, fast mode, parallel agents, and automations can consume more.
One timing note: do not base your buying decision on the old 2x boost promotion. OpenAI’s Codex help page says that promotion was specific to Pro tiers through May 31, 2026. Since today is June 9, 2026, treat it as over unless the live pricing page says otherwise.
Does the June 2026 Update Change the Buying Decision?
Yes, but only for some users.
If you are one person using Codex for code, the decision is still mostly Free vs Plus vs Pro. Start where you are, then upgrade only when limits block useful work.
If you are a founder, operator, analyst, marketer, or product lead, the June 2026 update changes the question. You are no longer asking only “Can Codex code?” You are asking whether Codex can create useful internal work products: dashboards, planners, review pages, documents, spreadsheets, campaign assets, or lightweight apps.
Use this rule:
| Situation | Better decision |
|---|---|
| You only want to test Codex | Stay on Free first |
| You build small projects weekly | Plus is the default |
| You hit limits during paid work | Try Pro before changing your whole stack |
| You need dashboards, Sites, plugins, and workspace controls | Evaluate Business or Enterprise |
| You want Codex inside your own product or automation | Use API pricing, not ChatGPT plan math |
The danger is upgrading because the product sounds powerful. The better approach is to run three real tasks first: one code task, one review task, and one work-product task. If Codex completes all three with less cleanup than your current workflow, the upgrade conversation becomes concrete.
How Much Does the Codex API Cost?
The Codex API is separate from ChatGPT plan access.
As of writing, OpenAI lists GPT-5.3-Codex API pricing at:
| Model | Input | Cached input | Output | Context |
|---|---|---|---|---|
| GPT-5.3-Codex | $1.75 per 1M tokens | $0.175 per 1M tokens | $14.00 per 1M tokens | 400K context |
Use API pricing if you are building your own product, internal coding workflow, CI automation, or custom agent system. Use ChatGPT plan pricing if you are using the official Codex surfaces.
The mistake is mixing them together. A ChatGPT Plus subscription does not mean you can run unlimited GPT-5.3-Codex API calls. API usage has its own billing and rate limits.
If you see older Codex model names in a third-party comparison, check OpenAI’s live pricing page before budgeting around them. The API page can change faster than blog roundups.
When Is ChatGPT Plus Enough for Codex?
ChatGPT Plus is enough when Codex is a productivity booster, not your full-time engineering workforce.
Plus usually makes sense if you:
- Build one or two projects at a time
- Use Codex for bug fixes and small features
- Want help understanding a codebase
- Ask Codex to write tests occasionally
- Need help with documentation, scripts, and small refactors
- Use Codex alongside tools like Cursor, GitHub Copilot, or Claude
For many solo users, Plus is the sweet spot. It gives you serious access without jumping straight to a power-user subscription.
If you are still learning, pair this with our Best AI Coding Tools 2026 guide so you can decide whether Codex should be your main agent or just one tool in the stack.
If you are not mostly coding, pair this with Codex for Knowledge Work 2026 instead. That guide looks at the non-developer side of Codex: reports, spreadsheets, dashboards, and internal tools.
When Is ChatGPT Pro Worth It for Codex?
ChatGPT Pro is worth considering when Codex starts saving paid hours.
Upgrade to Pro if at least two of these are true:
- You run Codex daily.
- You hand off multi-file changes often.
- You manage several tasks in parallel.
- You use the Codex app, CLI, and IDE extension together.
- You rely on Codex for review, tests, migrations, or release work.
- You hit limits during real work, not just experiments.
- One saved client hour covers most of the subscription cost.
Do not upgrade because Pro sounds more professional. Upgrade when limits are slowing down work that already creates value. If you are comparing Pro options, check whether you need the lower Pro path or the higher 20x usage path on OpenAI’s ChatGPT Pro tiers page.
The best test is simple: keep a one-week log. Write down every time Codex saved more than 20 minutes, and every time a limit blocked you. If that list is short, stay on Plus. If it is long, Pro is easy to justify.
Codex vs Claude Code vs GitHub Copilot Pricing
Codex is not the only paid coding choice. The real comparison is workflow, not sticker price.
| Tool | Pricing logic | Best for | Weak spot |
|---|---|---|---|
| OpenAI Codex | Included with ChatGPT plans plus credits/PAYG options | You already pay for ChatGPT or need multi-agent workflow | Can get confusing when plan limits, credits, and API billing overlap |
| Claude Code | Claude subscription/API access | You want terminal-native coding depth | Less attractive if your team is already standardized on ChatGPT |
| GitHub Copilot | Per-user coding subscription | Your team already lives in GitHub and VS Code | Less agentic than Codex or Claude Code for larger delegated tasks |
| Cursor | Editor subscription | You want the editor itself to be AI-native | Requires moving your daily editor workflow |
If you want one broad assistant plus a coding agent, Codex is attractive because it rides on your ChatGPT plan and now reaches more team workflows.
If you want deep terminal work, Claude Code remains a very strong pick.
If your company already lives in GitHub and VS Code, Copilot may be the easiest rollout.
If you want an AI-native editor, see Cursor vs Windsurf vs Zed 2026.
Before paying more, judge Codex from three finished tasks: one bug fix, one feature, and one code review. If it produces reviewable changes on all three and limits block useful work, upgrade. If it mainly feels impressive but does not finish work, stay on your current plan.
Final Recommendation
OpenAI Codex pricing in 2026 is better than it looks because you do not always need a separate coding-agent subscription.
Start with the Codex access included in your current ChatGPT plan. Use Free for testing, Plus for regular solo work, Pro for heavy daily use, and Business or Enterprise when team controls, Sites, plugins, and shared workflows matter.
Use the API only when you are building a custom system, not when you simply want to use Codex as a coding assistant.
The simplest buying rule:
- Free: Learn the workflow.
- Plus $20: Best default for individuals.
- Pro $100: Worth it when Codex supports real weekly projects.
- Pro $200: Worth it when Codex saves paid hours across heavy or parallel work.
- Business/Codex credits: Worth it when a team needs governed workspace access or higher Codex usage.
- Business/Enterprise: Worth it when Sites, plugins, admin controls, and shared workflows matter.
- API: Worth it when Codex becomes part of your own product or automation.
That keeps the decision grounded in work done, not AI-tool hype.
FAQ
Is OpenAI Codex free with ChatGPT?
Yes. OpenAI says Codex is included across eligible ChatGPT plans, including Free. Usage limits vary by plan, so free access is best for testing and small tasks.
Is Codex included with ChatGPT Plus?
Yes. ChatGPT Plus users can use Codex through supported Codex clients. Plus is the best default plan for many solo developers and technical creators.
Did Codex pricing change after the June 2026 update?
The plan logic did not become “everyone must upgrade.” The practical change is that Codex now has more workspace value for teams because of plugins, Sites, annotations, and non-developer workflows. Individuals should still start with existing access and upgrade only when usage limits block valuable work.
Do I need ChatGPT Pro for Codex?
Not at first. Use Plus until limits block valuable work. Pro makes sense when Codex saves enough time to justify the higher monthly cost.
Is Codex API pricing included in ChatGPT Plus?
No. API usage is billed separately from ChatGPT subscriptions. If you use GPT-5.3-Codex through the OpenAI API, token pricing and API rate limits apply.
What is Business Codex or Codex for business users?
For business users, Codex is tied to workspace access, admin controls, credits, and usage management. Business ChatGPT & Codex is the broader workspace path. Development-focused teams should also look at Codex credit options and the current rate card because actual cost depends on task size, model choice, fast mode, and parallel usage.
Is Codex cheaper than Claude Code?
It depends on your existing subscriptions. Codex can be cheaper if you already pay for ChatGPT and use the included access. Claude Code may be better value if your main work is terminal-native development and careful repo-level reasoning.
Related Reads on tossitt.com
- OpenAI Codex App Guide 2026
- Codex for Knowledge Work 2026
- Claude Code vs OpenAI Codex 2026
- Claude Code Guide 2026
- Best AI Coding Tools 2026
- Cursor vs Windsurf vs Zed 2026
