OpenAI Codex Pricing: Plans, Credits, and Real Cost
What Codex actually costs, from the Free plan to Enterprise, and why your bill can swing.
OpenAI Codex does not have its own price tag. It is bundled into every ChatGPT plan, so your plan choice sets your cost.
Plans run from $0 (Free) to $200/mo (Pro 20x), plus per-user Business seats and custom Enterprise deals. Heavy use is metered in credits.
This guide breaks down each tier, the April 2026 credit shift, and the real per-developer spend most teams see.
Is OpenAI Codex free?
Yes, OpenAI Codex is free on the ChatGPT Free plan, but access is limited. You get a small amount of usage before you hit a cap.
There is no separate Codex product to buy. Codex ships inside ChatGPT, so a paid plan simply unlocks more Codex usage.
For light testing, Free works. For daily coding, most developers upgrade to a paid tier for more capacity and cloud features.
Want help sizing your Codex plan and forecasting credit spend before you commit? Layer3 Labs builds the budget with you.
Book a ConsultationCodex pricing by ChatGPT plan
Codex pricing follows the ChatGPT plan ladder. Each tier adds more usage, more features, and better team controls.
Here is how the tiers compare, from free access to custom Enterprise deals.
- Free — $0/mo. Limited Codex access. Best for trying Codex or very light use.
- Go — $8/mo. More usage than Free. Fits solo hobbyists and students.
- Plus — $20/mo. Solid daily access. Fits individual developers doing regular work.
- Pro — $100/mo (Pro 20x at $200/mo). Much higher limits. Fits power users running heavy or parallel workloads.
- Business — about $20/user/mo billed annually ($25/user/mo monthly). Adds cloud features, GitHub, Slack, and Linear integrations, plus admin controls. Fits teams.
- Enterprise — custom pricing. Shared credit pool, no fixed rate limits, and dedicated support. Contact OpenAI sales.
The April 2026 shift to token-aligned credits
On April 2, 2026, OpenAI changed how Codex usage is measured. It moved from per-message limits to credits that track API tokens.
This means your usage is now metered like API calls. Bigger prompts, larger models, and more agents all burn more credits.
The change hit new and existing Plus, Pro, and Business plans, plus new Enterprise. Your seat price is the same, but usage now scales with real work.
Codex API and rate-card pricing for programmatic use
Codex also has a rate card for programmatic use, priced like API tokens. This covers automated jobs that run Codex outside the chat window.
Codex runs on the GPT-5.6 model family. Token-aligned credits let heavy or scripted use meter cleanly against that rate card.
If you wire Codex into pipelines or bots, budget by tokens, not by seats. Check the current rate card before you scale automation.
What Codex actually costs per developer per month
In practice, Codex runs about $100 to $200 per developer per month on average. That is real usage, not just the sticker seat price.
The range is wide because usage varies. Model choice, parallel agents, automations, and fast mode all push the number up or down.
A light user may stay near the seat price. A power user running many cloud agents can land at the top of the range or higher.
How to control Codex costs
The single biggest cost driver is credit overage, not seats. Because Codex meters token-aligned credits, heavy use can blow past a plan's included amount fast.
One developer running several parallel cloud agents or heavy fast mode can burn through credits in days, not weeks.
Budget for overages, watch your usage dashboard, and set team norms so spend stays predictable.
- Match the model to the task. Reserve the biggest models for hard problems.
- Limit parallel cloud agents unless the job truly needs them.
- Use fast mode sparingly. It is convenient but credit-hungry.
- Track usage weekly so a surprise bill never lands.
- Right-size seats. Move heavy users to Pro instead of over-buying credits.
Codex pricing vs Claude Code, Cursor, and Copilot
Codex is bundled into ChatGPT plans, so it has no standalone price. Claude Code, Cursor, and GitHub Copilot price their coding tools differently.
The table below shows the headline structure. Exact numbers change often, so check each tool before you commit.
For deeper breakdowns, see our Claude Code pricing guide and Cursor pricing guide.
- Codex — no standalone price; included in ChatGPT plans ($0 to $200/mo, plus Business seats). Metered in credits.
- Claude Code — priced through Anthropic plans and API usage. See our pricing guide for current tiers.
- Cursor — its own subscription tiers with usage-based add-ons. See our pricing guide for current tiers.
- GitHub Copilot — flat per-user subscription with business and enterprise seats.
Is Codex worth it?
Codex is worth it for most active developers, because a $20 to $100 plan often pays back in saved hours within days.
Solo developers do well on Plus or Pro. Teams get more value from Business, thanks to shared admin and integrations.
Enterprises with heavy automation should model credit usage first. A shared credit pool and custom terms can beat per-seat math at scale.
Frequently Asked Questions
- Yes, Codex is free on the ChatGPT Free plan, but access is limited. You get a small amount of usage before hitting a cap. Daily coders usually upgrade to a paid plan for more capacity.
- Codex has no separate price. It is included in ChatGPT plans that run from $0 (Free) to $200/mo (Pro 20x), with Business seats around $20 per user per month on annual billing and custom Enterprise pricing.
- No. Codex ships inside ChatGPT, so your ChatGPT plan is your Codex plan. Upgrading your ChatGPT tier is how you unlock more Codex usage.
- On April 2, 2026, OpenAI moved Codex usage from per-message limits to token-aligned credits. Usage is now metered like API tokens, so busier weeks cost more even though your seat price stays the same.
- Most teams see about $100 to $200 per developer per month in real usage. The range depends on model choice, parallel agents, automations, and fast mode. The plan price is a floor, not a ceiling.
- Budget for credit overages, not just seats. Match models to tasks, limit parallel cloud agents, use fast mode sparingly, and track usage weekly. Overages, not the plan tier, cause most surprise bills.
- It depends on usage. Codex bundles into a ChatGPT plan, so you may already own it. Claude Code and Cursor price separately. Compare our [Claude Code pricing](/guides/claude-code-pricing) and [Cursor pricing](/guides/cursor-pricing) guides for the current numbers.
Not sure which Codex plan fits your team?
Layer3 Labs helps teams pick and budget agentic coding tools. We size your plan, model your credit spend, and set guardrails so costs stay predictable.
Book a Consultation