OpenAI Codex Pricing: Complete Guide to Plans, Costs, and Hidden Fees (2026)

OpenAI Codex is bundled into ChatGPT subscriptions. ChatGPT Plus costs $20 per month for light weekly use. Pro 5x costs $100 per month for daily work. Pro 20x costs $200 per month for heavy parallel tasks. Business and Enterprise plans serve teams needing SSO admin controls and flexible credits. API access using codex-mini-latest costs $1.50 per million input tokens and $6.00 per million output tokens. Real-world costs average $100 to $200 per developer per month depending on model choice, Fast mode, and automation volume.

You know what Codex does. You need to know what it costs and which plan will not surprise you at month-end. This guide translates OpenAI’s credit and token system into plain monthly budgets by persona. It exposes hidden cost multipliers and gives you a 60-second decision matrix. We cover ChatGPT tiers from Free to Enterprise, API pricing, credit math, hidden costs, team buying, and GitHub Copilot comparison. No feature fluff.

What are the current OpenAI Codex pricing tiers?

ChatGPT Plus at $20 per month includes Codex Web, CLI, and IDE with 30 to 150 messages per 5-hour window. Pro 5x at $100 per month gives five times those limits. Pro 20x at $200 per month gives twenty times the limits. Business at $25 to $30 per seat per month adds SSO, shared credit pool, and no training on your data. Enterprise offers custom pricing with no fixed limits, SCIM, EKM, RBAC, audit logs, and data residency. Free and Go tiers provide local-only access with no cloud Codex features.

Which ChatGPT plan is right for my Codex usage?

Solo developers running a few sessions per week should choose Plus at $20 per month. Developers coding daily with parallel tasks should choose Pro 5x at $100 per month. Engineers running heavy automation or multiple agents should choose Pro 20x at $200 per month or use the API directly. Teams of five or more needing single sign-on and admin controls should choose Business. Organizations with compliance requirements or needing data residency should choose Enterprise.

Choose This If…

Choose Plus ($20/mo) if:

• You code a few sessions per week
• You work on one task at a time
• You want predictable monthly cost
• You do not need admin controls

Choose Pro 5x ($100/mo) if:

• You use Codex daily
• You run multiple tasks in parallel
• You hit Plus limits regularly
• You need more Fast mode headroom

Choose Pro 20x ($200/mo) if:

• You run heavy automation workflows
• You operate many parallel agents
• You need maximum Fast mode capacity
• You build CI/CD pipelines with Codex

Choose Business ($25-30/seat) if:

• You manage a team of 5+ developers
• You need SAML SSO and MFA
• You want a shared credit pool
• You require zero data training on your code

Choose Enterprise if:

• You need SCIM provisioning
• You require EKM and RBAC
• You need audit logs and data residency
• You have compliance requirements (HIPAA, SOC 2)

How much does OpenAI Codex cost per month in practice?

OpenAI states average Codex costs run $100 to $200 per developer per month. Solo developers on Plus spend roughly $20 per month since usage is included. Heavy Pro users spend $100 to $200 plus any overage credits. Automation via API ranges from $50 to $500 per month depending on volume. Fast mode and image generation can burn credits three to five times faster than standard mode.

How do Codex credits and tokens work?

Credits are the unified currency for Codex and other agentic features. The rate card charges credits per one million tokens by type: input tokens, cached input tokens, and output tokens. GPT-5.6 Sol costs 125 credits per million input tokens, 12.5 credits per million cached input tokens, and 750 credits per million output tokens. GPT-5.6 Terra costs 62.5, 6.25, and 375 credits respectively. GPT-5.6 Luna costs 25, 2.5, and 150 credits respectively. A typical Codex task consumes 5 to 45 credits. Fast mode increases the credit burn rate.

Simple Version

Think of credits like a prepaid phone card. Each model has a price per “unit of work” measured in tokens. Simpler models (Luna) cost fewer credits. Powerful models (Sol) cost more credits. Output tokens cost the most. Cached input tokens cost the least. Your plan gives you a monthly credit allowance. Going over means buying more credits.

What hidden costs should I watch for with Codex?

Fast mode burns credits two to three times faster than standard mode. Image generation uses limits three to five times faster than text tasks. MCP servers add context which means more input tokens and higher credit consumption. GitHub code reviews have separate rate limits that can catch you off guard. Auto-reload on credit depletion can lead to surprise charges if not monitored.

Watch For These Cost Traps

Fast Mode

• Burns credits 2-3x faster
• Great for speed, expensive for volume
• Turn off for batch tasks

Image Generation

• Uses message limits 3-5x faster
• Each generation counts heavily
• Limit use in automated workflows

MCP Servers

• Extra context = more input tokens
• Each tool call adds tokens
• Audit your MCP usage weekly

Code Review (GitHub Integration)

• Separate rate limits apply
• Can consume credits independently
• Monitor if you review many PRs

Auto-Reload

• Automatically buys credits when low
• Convenient but can surprise you
• Set spending alerts in billing dashboard

Should I use the ChatGPT subscription or the OpenAI API for Codex automation?

Subscription plans are cheaper when you use near your included allowance because the monthly fee covers a bundle of credits. The API is cheaper for sporadic or very high-volume automation where you only pay for what you use. The codex-mini-latest model via API costs $1.50 per million input tokens and $6.00 per million output tokens. Break-even typically occurs around 200 tasks per month. For CI/CD pipelines running daily, API often wins. For interactive daily coding, subscription wins.

Decision Formula

Use Subscription (Plus/Pro/Business) if:

• You code interactively most days
• You stay within 5-hour window limits
• You want predictable monthly cost
• You need IDE integration and web UI

Use API (codex-mini-latest) if:

• You run automated CI/CD jobs
• Usage is bursty or unpredictable
• You want per-token billing
• You build custom Codex agents

Break-even calculation:

Monthly API cost = (Input tokens × $1.50 + Output tokens × $6.00) / 1,000,000
Compare to: Pro 5x ($100) or Pro 20x ($200) monthly fee

Compare to: Pro 5x ($100) or Pro 20x ($200) monthly fee

How does OpenAI Codex pricing work for teams?

ChatGPT Business costs $25 per seat per month billed monthly or $30 per seat per month billed annually. Each seat gets Plus-level limits. Credits are pooled across the workspace. Admins get SAML SSO, MFA enforcement, and zero data training by default. Enterprise removes fixed limits and adds SCIM provisioning, EKM, RBAC, audit logs, and data residency options. Minimum seat counts typically start at five seats for Business.

Upgrade Triggers

Move from Plus to Business when:

• You have 5+ developers using Codex
• You need SSO for security policy
• You want centralized billing
• You need admin usage visibility

Move from Business to Enterprise when:

• You need SCIM auto-provisioning
• You require EKM (Enterprise Key Management)
• You need RBAC for different roles
• You have data residency requirements
• You need HIPAA compliance

Is Codex cheaper than GitHub Copilot?

GitHub Copilot charges a flat $19 per month for Individuals and $39 per month per seat for Business. This is predictable but limited to inline completions and chat. OpenAI Codex ranges from $20 to $200+ per month with variable costs based on usage. Codex handles complex multi-file refactors, code reviews, and automation that Copilot cannot. Choose Copilot for predictable budget and inline completions. Choose Codex for powerful refactors, reviews, and multi-step automation.

Choose Codex if: You need multi-file refactoring, automated code reviews, CI/CD integration, or agentic workflows.

Choose Copilot if: You want predictable per-seat pricing, mostly need inline completions, and prefer GitHub-native experience.

Why did my Codex costs change in 2026?

On April 2, 2026, OpenAI migrated from per-message billing to token-based credit billing. The legacy system charged roughly 14 to 34 credits per message regardless of length. The new system charges by token type: input, cached input, and output. Impact varies. Output-heavy tasks and Fast mode now cost more. Input-heavy tasks with cached context cost less. Most workspaces have migrated. Check your workspace migration status in settings if you are unsure.

Before vs After

Final Recommendation

Start with ChatGPT Plus at $20 per month if you are a solo developer. Upgrade to Pro 5x at $100 per month when you hit daily limits. Move to Pro 20x at $200 per month or API for heavy automation. Teams of five or more should evaluate Business at $25 to $30 per seat for SSO and shared credits. Enterprise is for compliance and scale. Monitor Fast mode, image generation, and MCP usage to avoid bill shock. Use the OpenAI Agents SDK for CI/CD pipelines. Compare against GitHub Copilot if you only need inline completions and want fixed pricing.

FAQ

Is Codex free with ChatGPT Plus?

Yes. Codex Web, CLI, IDE, and local chats are included in ChatGPT Plus at $20 per month with usage limits of 30 to 150 messages per 5-hour window depending on the model. No separate Codex subscription exists.

How many Codex messages do I get per month?

There is no fixed monthly cap. Limits are rolling 5-hour windows. For example, Plus users get 20 to 110 Terra model messages per 5-hour window. Weekly caps may also apply. Heavy users hit limits daily. Pro tiers multiply these windows 5x or 20x.

What happens when I hit my Codex limit?

Active tasks complete. You can buy extra credits on Plus or Pro, upgrade to a higher tier, switch to a smaller model like Luna, or wait for the 5-hour window to reset. Business and Enterprise workspaces can purchase shared workspace credits.

Can I use Codex via API without a ChatGPT subscription?

Yes. Use codex-mini-latest or GPT-5 models via OpenAI API key. Pay per token at $1.50 per million input tokens and $6.00 per million output tokens for codex-mini-latest. No ChatGPT plan required. API usage is separate from subscription credits.

What are Codex credits and how do they work?

Credits are the unified currency for Codex and agentic features. They are consumed per token by model type. Rate card example: GPT-5.6 Sol charges 125 credits per million input tokens, 12.5 credits per million cached input tokens, and 750 credits per million output tokens. Your plan includes a credit allowance per window. Overage draws from purchased credits.

Does Fast mode cost extra?

Yes. Fast mode consumes credits at a higher rate for supported models. It is not a separate fee but burns your allowance two to three times faster. Check the speed documentation for model-specific rates.