API boundary
ChatGPT plan access, Codex credits, and OpenAI API token billing are separate systems. API usage should be modeled from OpenAI API pricing, not ChatGPT plan allowances.
Pricing
Codex Pricing 2026: GPT-5.6 Plans, Credits, and API Costs
Codex is included with ChatGPT Free, Go, Plus, Pro, Business, Edu, and Enterprise. ChatGPT Work shares Codex usage, while API-key workflows use separate token billing.
AI Coding Assistants
Pricing checked July 9, 2026
Buyer guide
Where to start before you compare plans
Check who bills you, how usage is metered, and which caveats apply before choosing an access path.
Tracks
Which plan fits whom
ChatGPT plan access
Use the ChatGPT route when you want first-party Codex surfaces and plan-based usage limits instead of a separate developer-platform bill.
Best for: Solo developers and teams already choosing a ChatGPT subscription tier.
Avoid if: You need API keys, CI automation, or a separately metered platform budget.
Codex usage credits
Use the current token-based Codex rate card when estimating how GPT-5.6 work consumes credits after included plan limits.
Best for: Buyers modeling heavier Work or Codex sessions, parallel tasks, max reasoning, or ultra multi-agent runs.
Avoid if: You only need a simple yes/no answer about whether your ChatGPT plan unlocks Codex.
Business workspace
Use Business when admin controls, workspace credits, and existing Codex-seat eligibility matter more than a single individual subscription tier. New Business workspaces should not assume they can add Codex-only seats.
Best for: Teams using standard Business ChatGPT seats, existing eligible Codex-only seats, or Enterprise/Edu flexible pricing conversations.
Avoid if: You are a solo user without workspace governance needs.
API route
Use API pricing when you are building custom coding-agent systems, product integrations, or CI workflows with API keys.
Best for: Engineering teams with programmatic automation and separate platform spend.
Avoid if: You mainly want human-driven repository work inside Codex product surfaces.
Access paths
Subscription, API, and workspace routes
Each access path shows who owns the bill and whether access is bundled, separately metered, sold as an add-on, or handled through sales.
Bundled appShared subscription quotaRecommended route
ChatGPT plan access
Codex is included across current ChatGPT plans on desktop, web, CLI, and IDE surfaces; ChatGPT Work follows the same usage structure.
Best for: Developers who want subscription-backed GPT-5.6 agent usage instead of a separate API bill.
Boundary: Free and Go use GPT-5.6 Terra; Plus, Pro, Business, and Enterprise have broader model choice and higher limits. Legacy Codex-only Business seats require a seat before June 24, 2026 or a pending invite as of June 24.
Open ChatGPT pricing contextBundled appShared subscription quota
ChatGPT Work and Codex surfaces
Use first-party ChatGPT Work and Codex surfaces for interactive agent work with shared plan limits and credits.
Best for: Interactive repository work, code review, and agent sessions managed by OpenAI product UX.
Boundary: This route is plan- and credit-based. Work and Codex share the agentic usage pool; API-key usage remains separate.
Open Codex pricing contextDirect APISeparate API meter
GPT-5.6 API models
Use OpenAI API pricing for programmatic systems built with gpt-5.6-sol, gpt-5.6-terra, or gpt-5.6-luna.
Best for: Programmatic coding agents, internal tools, and product integrations.
Boundary: API spend is separate from ChatGPT subscription allowances and should be modeled by tokens or request units.
Open Codex pricing contextWatchouts
What buyers often miss
These are the boundary conditions and purchase traps worth checking before you optimize for the lowest headline number.
ChatGPT prices belong on ChatGPT pricing
Codex can be reached through ChatGPT plans, but this page should not duplicate the full ChatGPT plan table. Open the canonical ChatGPT pricing page for exact subscription tiers.
Credits are not the same as API spend
Buying or consuming Codex credits extends plan-based Codex usage. It does not prepay or discount OpenAI API token bills.
Business Codex-only seats are eligibility-bound
Standard ChatGPT Business seats include Codex. A workspace can manage legacy Codex-only seats only if it had a seat before June 24, 2026 or a pending invite as of June 24; new workspaces cannot add a first Codex-only seat.
The rate card is the Codex-owned fact layer
The current dedicated rate card is token-based and says GPT-5.6 usage averages 5–40 credits per message. Treat that as an estimate and use the current rate card for the authoritative token mix.
Editorial pricing notes
Pricing notes
Plan caveats, contract terms, and feature-access limits that can change what you actually pay.
How Codex pricing works in 2026
Codex pricing has three main budget paths: included usage through a ChatGPT plan or workspace, credits that extend eligible ChatGPT Work and Codex usage, and separately billed API-key token usage. Legacy Codex-only Business seats remain relevant only to workspaces with a seat before June 24, 2026 or a pending invite as of June 24.
Is Codex included with ChatGPT?
Yes. Codex is included across current ChatGPT plans. ChatGPT Work and Codex draw from the same pricing, credits, and usage-limit structure, while API-key usage remains separate.
Codex credits vs OpenAI API billing
Codex credits are a token-based product usage meter shared with eligible ChatGPT Work and other agentic features. They do not pay OpenAI API invoices. Plus and Pro use included allowance first, then purchased credits where eligible; API keys use separate developer billing.
Best Codex plan by use case
Free offers limited entry access; Go is the $8 paid entry where local pricing applies; Plus is $20; Pro starts at $100 for 5x usage and offers a $200 20x tier. Business is $25 per user monthly or $20 per user monthly on annual billing with at least two users. These are ChatGPT product prices, not API token rates.
Buying path
Start with ChatGPT plan access when work runs through ChatGPT desktop, web, CLI, IDE, or a managed workspace. Free and Go use GPT-5.6 Terra; Plus, Pro, Business, and Enterprise can choose Sol, Terra, or Luna. Use API billing when a key, CI workflow, or custom system owns the work.
Upgrade triggers
Upgrade when coding work repeatedly runs into capacity, credit, or collaboration limits that block real delivery. A higher lane makes sense when Codex is used across multiple repositories, long-running tasks, or team workflows rather than occasional experiments.
API and team boundaries
Keep shared ChatGPT Work/Codex usage, credits, Business workspace access, and OpenAI API billing separate. Legacy Codex-only Business seats remain available only to workspaces with a seat before June 24, 2026 or a pending invite as of June 24; they are not a normal option for new buyers.
Final pricing check
Before committing, verify which Codex surface the team will use, how credits or limits are consumed, and whether the buyer needs workspace governance or token-priced model access. If the main job is programmatic integration, price the API route directly instead of treating a seat as the budget.
Evidence boundary
Official sources
Only explicitly official evidence is listed here.
FAQ
Codex pricing FAQ
What is the cheapest paid plan that makes Codex feel usable?
Go at $8 per month in the US is the cheapest paid Codex entry, but it should be treated as limited access. Plus at $20 per month is the first dependable self-serve tier for focused weekly coding sessions.
Do Pro tiers change Codex quality or mostly usage?
Plus and Pro can both use the GPT-5.6 Sol, Terra, and Luna family and Codex ultra. Pro primarily adds 5x or 20x usage headroom and may include separate research-preview access; it is not required merely to reach the GPT-5.6 family.
Can new Business workspaces add Codex-only seats?
No. Only a Business workspace that had a Codex seat before June 24, 2026 or a pending invite as of June 24 can continue managing or adding Codex-only seats. New workspaces cannot add a first Codex-only seat.
Do my ChatGPT payments cover API use?
No. ChatGPT plan access and API token billing are separate systems. If you use Codex with an API key or build your own workflows, you should expect a separate API bill.
What happens after I hit included Codex limits?
OpenAI lets eligible plans extend Codex usage with credits. That keeps you on the plan-based Codex path, which is still different from paying API token rates directly.
Are Codex API calls included in ChatGPT?
No. ChatGPT subscription access and OpenAI API billing are separate. If a workflow uses an API key, model that cost from the OpenAI API pricing path rather than the ChatGPT seat price.
What is the best Codex plan for code review?
For individual review work, start with the lowest ChatGPT plan that provides enough usage. For teams, compare standard Business controls and workspace credits; only workspaces with a seat before June 24, 2026 or a pending invite as of June 24 should consider legacy Codex-only seats. API-based automation remains separate.
Internal links
What to open next
Keep researching Codex
Pair the pricing snapshot with verdict, alternatives, and the full profile page.
ToolProfile: CodexOpenAI's coding agent with GPT-5.6 model choice, selectable reasoning, and parallel Ultra workflows.Review: Codex Review: Is GPT-5.6 Worth It for Developers?Codex pairs GPT-5.6 model choice with selectable reasoning and Ultra subagents, but its value depends on managing plan limits, credits, surfaces, and separate API billing.Alternatives: Best Codex Alternatives in 2026Cursor, GitHub Copilot, and Windsurf are the clearest Codex alternatives if you want a more editor-first workflow, cheaper seat pricing, or a different balance between agent power and IDE convenience.
Head-to-head pages
Open direct comparison pages before choosing a plan.
VSCodex vs Claude CodeThis comparison is a real tie: Claude Code is stronger for local-first coding workflows and deployment flexibility, while Codex is stronger for cloud orchestration and supervisory multi-agent work.VSCodex vs CursorCursor is the better default for most developers because it combines an AI-native editor, broader model choice, and stronger day-to-day workflow tooling. Codex is stronger for multi-agent delegation, cross-tool reach, and direct OpenAI API usage.VSCodex vs GitHub CopilotCodex should be the default when the team is buying for deeper agentic coding work: repository understanding, task execution, command-running, tests, code review preparation, and reviewable diffs. GitHub Copilot should lead when the team values GitHub-native IDE assistance, pull request workflow, seat administration, and organization policy more than standalone agent depth.VSCodex vs Devin Desktop (Windsurf)Codex is the preferred default when the buyer wants an autonomous coding agent that can handle repo-level tasks, run commands, produce diffs, support cloud delegation, and plug into review workflows. Windsurf is the better switch when the real purchase is an AI-native IDE with autocomplete, Cascade, previews, local context awareness, model routing, and a tighter in-editor flow.
Other reviews in this category
Sanity-check nearby tools before committing to a pricing tier.
ReviewCursor ReviewCursor is one of the strongest AI coding environments available, pairing deep codebase context with agentic workflows, but its usage-based pricing makes plan selection less predictable.ReviewGitHub Copilot ReviewGitHub Copilot remains one of the easiest AI coding assistants to adopt, combining GitHub-native workflow fit, broad IDE coverage, code review, and flexible plans from free to enterprise.ReviewClaude Code ReviewClaude Code remains a top coding agent for serious repo work. Fable 5 affects Claude-side evaluation and budget planning, but do not assume it is the default Claude Code model without official confirmation.ReviewWindsurf ReviewWindsurf is one of the strongest AI-native IDEs available, with multi-file agent workflows, MCP, browser previews, and JetBrains support. Its main tradeoffs are hybrid overage pricing and a less mature ecosystem than safer editor-plus-plugin setups.