OpenAI Codex Pricing 2026: Plans, Credits, Rate Card, and Usage Limits Explained

Pricing last checked: June 2026

Quick Answer: How OpenAI Codex Pricing Works in 2026

OpenAI Codex does not have one standalone monthly price. In 2026, access depends on your ChatGPT plan, workspace plan, or API key. Included usage is limited, and extra usage is priced through credits that map to token usage: input tokens, cached input tokens, and output tokens. Real monthly cost varies by model, task size, Fast mode, automations, and how many Codex tasks a developer runs.

In practice, Codex pricing falls into three buckets: included usage inside ChatGPT plans, additional credit-based usage after limits, and API-key usage billed at standard OpenAI API rates. OpenAI lists Codex as included in ChatGPT Free, Go, Plus, Pro, Business, Edu, and Enterprise plans, while API-key usage is separate and billed through the OpenAI Platform.

Codex Pricing at a Glance

OpenAI’s Codex pricing page lists Free, Go, Plus, Pro, Business, Enterprise, Edu, and API-key options. The same page says Plus includes Codex on the web, CLI, IDE extension, and iOS, while Pro gives either 5x or 20x higher Codex usage than Plus. OpenAI’s Pro help page clarifies that Pro $100 unlocks 5x higher usage than Plus, while Pro $200 unlocks 20x higher usage than Plus.

OpenAI Codex Plans and Access Options

Free and Go

Free and Go are entry-level ways to try Codex. Free is best for testing Codex capabilities on quick coding tasks, while Go is positioned for lightweight coding tasks at $8/month. These plans are useful for learning the workflow, but they are not the best fit for regular development work, larger repositories, or repeated agentic coding sessions.

Free and Go users should treat Codex as a trial or occasional assistant rather than a predictable development budget. OpenAI’s credits documentation says Free and Go users are prompted to upgrade to Plus instead of adding Codex credits when they hit Codex limits.

Plus

ChatGPT Plus is the first practical plan for many individual developers. It costs $20/month and includes expanded Codex usage, access through the Codex web app, CLI, IDE extension, and iOS, plus cloud-based integrations such as automatic code review and Slack integration where available.

Plus is enough if you use Codex for a few focused coding sessions each week: debugging a function, generating tests, explaining unfamiliar code, writing small features, or reviewing a contained change. It is not the safest budget choice if Codex is part of your daily engineering workflow.

Pro 5x and Pro 20x

Pro is for heavier individual use. OpenAI lists Pro as starting at $100/month, with 5x or 20x more Codex usage than Plus. OpenAI’s Pro tier documentation says the $100 Pro tier gives 5x higher usage than Plus, while the $200 Pro tier gives 20x higher usage than Plus.

Choose Pro 5x if Codex is part of your regular work but not running constantly. Choose Pro 20x if you rely on Codex throughout the day for multi-step implementation, larger refactors, repeated code review, and long-running coding sessions.

Business

ChatGPT Business supports standard ChatGPT seats and Codex-only seats. Standard ChatGPT seats include ChatGPT and Codex access and are billed at a fixed per-user price, while Codex-only seats provide access to Codex only and are usage-based with no fixed monthly seat price.

For teams, this matters because not every user needs the same seat type. A product manager, analyst, or operator may need a standard ChatGPT seat, while a build agent or engineering workflow may only need Codex access. Business workspaces can also purchase workspace credits to extend usage beyond included limits.

Enterprise and Edu

Enterprise and Edu plans are designed for organizations that need security controls, governance, analytics, and flexible usage. OpenAI’s Codex pricing page describes Enterprise and Edu as custom plans with enterprise-grade functionality, including security controls, audit logs, data retention controls, and usage monitoring.

For Enterprise and Edu users with flexible pricing, OpenAI says there are no fixed Codex rate limits because usage scales with credits. Enterprise and Edu plans without flexible pricing generally follow the same per-seat usage limits as Plus for most features.

API Key

API-key usage is different from ChatGPT-plan usage. When you sign in to Codex with an API key, OpenAI bills usage through your OpenAI Platform account at standard API rates, not through your included ChatGPT plan credits.

API-key usage is best for local workflows, CI/CD jobs, programmatic Codex CLI workflows, SDK usage, and automation in shared engineering environments. It does not include every ChatGPT workspace or cloud feature; OpenAI notes that some features relying on ChatGPT workspace access or cloud services may be limited or unavailable with API-key authentication.

What Changed in April 2026?

The biggest Codex pricing change in 2026 was the move from message-based pricing to token-based credit pricing.

OpenAI’s Codex rate-card documentation says the new model replaces average per-message estimates with a direct mapping between token usage and credits. That means Codex cost is now tied to input tokens, cached input tokens, and output tokens rather than a simple “one message equals one fixed cost” model.

This is why older Codex pricing articles are often misleading. A small bug fix and a large refactor may both look like “one Codex task,” but they can consume very different amounts of context, reasoning, tool use, and generated code.

How Codex Credits Work

Codex credits are a flexible usage layer. Your included plan usage is consumed first. After you reach plan limits, eligible Plus and Pro users can use credits to keep working, while Business, Enterprise, and Edu workspaces can use shared workspace credits.

Codex credit usage is based on three token types:

The most important takeaway is that credit cost is not just about how many prompts you send. A short prompt that asks Codex to inspect a large repository, generate multiple files, run tools, and produce a long explanation can consume more credits than several small, scoped edits.

Current OpenAI Codex Rate Card

Pricing last checked: June 2026.

OpenAI’s current Codex rate card prices Codex usage in credits per 1 million tokens. The table below summarizes the main token-based rows relevant to Codex users.

OpenAI also lists GPT-Image-2.0 rows in the Codex rate card and notes that Fast mode consumes credits at a higher rate for supported models. OpenAI says a typical Codex task using GPT-5.5 may consume between 5 and 45 credits, though actual cost varies by model, number of running instances, automations, and Fast mode usage.

Codex Usage Limits Explained

Codex usage limits are not fixed message counts. OpenAI says the number of Codex messages you can send depends on the model used, the size and complexity of coding tasks, and whether work runs locally or in the cloud. Small scripts may consume only a fraction of your allowance, while larger codebases or long-running sessions use more per message.

For Plus, Pro 5x, Pro 20x, and Business, OpenAI publishes five-hour local-message usage ranges by model. These are ranges because actual usage depends on token consumption and task complexity.

OpenAI notes that local messages and cloud tasks share a five-hour window and that additional weekly limits may apply. It also says Enterprise and Edu users with flexible pricing have no fixed rate limits because usage scales with credits.

What Happens When You Hit Codex Limits?

If you hit Codex usage limits during an active turn, OpenAI says the agent can continue working on that turn, subject to fair use limits. After that, your options depend on your plan.

Plus and Pro users who reach their usage limit may be able to purchase additional credits. Business, Edu, and Enterprise workspaces with flexible pricing can purchase workspace credits. Users can also switch to a smaller model to stretch remaining limits, and all users may run extra local tasks with an API key billed at standard API rates.

For Free and Go users, OpenAI’s credits documentation says Codex users are prompted to upgrade to Plus instead of adding credits.

Realistic Monthly Cost Scenarios

These examples are planning scenarios, not guaranteed bills. Codex cost depends heavily on model choice, task size, generated output, automations, Fast mode, and how many tasks each developer runs. OpenAI itself says Codex averages around $100–$200 per developer per month, with large variance depending on model, number of running instances, automations, and Fast mode usage.

A practical budgeting rule is to start with the smallest plan that matches the workflow, then monitor real usage for two to four weeks. If developers frequently hit limits, upgrade the plan or add credits. If usage is mostly automated or CI-driven, API-key billing may be easier to forecast from token logs.

Codex API-Key Pricing vs ChatGPT-Plan Usage

ChatGPT-plan usage and API-key usage are separate.

When you use Codex through a ChatGPT plan, you use the included Codex allowance for that plan first, then credits if your plan or workspace supports them. When you use Codex with an API key, usage is billed through your OpenAI Platform account at standard API rates.

OpenAI’s API pricing page lists the following standard rates for the main coding-relevant models:

Use ChatGPT-plan access when you want Codex inside the app, CLI, IDE extension, and supported workspace features. Use API-key access when you want local automation, CI/CD jobs, scripts, or SDK-based usage where token-level billing is more natural.

How to Reduce Codex Credit Usage

The best way to reduce Codex cost is to reduce unnecessary context and avoid expensive modes when they are not needed. OpenAI recommends controlling prompt size, keeping instructions precise, reducing the size of AGENTS.md, limiting MCP servers, and switching to smaller models such as GPT-5.4 mini for routine tasks.

Practical ways to make Codex usage last longer:

• Scope each task tightly: ask for one fix, one file, or one behavior change instead of “review the whole app.”
• Use smaller models for routine edits, formatting, tests, and simple refactors.
• Keep AGENTS.md useful but not bloated; encode real build, test, lint, and PR rules, not vague preferences.
• Limit MCP servers when they are not needed because every extra tool can add context.
• Avoid Fast mode unless speed is worth the higher credit rate.
• Ask Codex to plan first for complex tasks so it does not waste work on the wrong implementation path.

OpenAI’s Fast mode documentation says Fast mode increases supported model speed but consumes credits at a higher rate: 2.5x the Standard rate for GPT-5.5 and 2x the Standard rate for GPT-5.4.

Codex vs Cursor, Claude Code, GitHub Copilot, and UI Bakery

Codex is best understood as an OpenAI coding agent. It is useful when you want code generation, code review, debugging, tests, refactoring, and agentic coding tasks connected to OpenAI models and ChatGPT plan access.

The comparison should not be framed as “one tool replaces everything.” Most teams choose based on where the work happens.

UI Bakery is not an AI coding editor and should not be positioned as a Codex replacement. If the end goal is an internal dashboard, admin panel, approval queue, or CRUD app around existing systems, UI Bakery may reduce the amount of custom code a team asks Codex to write and maintain. Codex is best for code work; UI Bakery is best for building and operating internal app interfaces.

Relevant internal links for this section:

• Cursor AI guide
• Cursor AI pricing
• AI app generator
• internal tools platform

Which Codex Plan Should You Choose?

Start with Plus if you are unsure. Move to Pro 5x when Codex becomes part of your normal development process. Move to Pro 20x when you regularly run into limits or use Codex across long, daily coding sessions. For teams, Business or Enterprise usually makes more sense than reimbursing many individual subscriptions.