CODEX PRICING AT A GLANCE — JUNE 2026
● Free: Codex included, very limited (try quick tasks only)
● Plus $20/mo: 15-80 GPT-5.5 cloud tasks per 5-hr window — best value for daily developers
● Pro 5x $100/mo: 50-300 tasks per 5-hr window — added April 9, 2026 to compete with Claude Code $100
● Pro 20x $200/mo: 200-1,200 tasks per 5-hr window — full-time Codex power users
● API: gpt-5.3-codex at $1.75/$14.00/M tokens — for CI/CD and products, not daily dev use
● Key change April 2, 2026: Moved from per-message to token-based credit billing — costs now variable by task size
● Typical session cost: $0.50-$2.00 at API rates. GPT-5.5 burns 5-45 credits per message
The April 2026 Pricing Overhaul — What Changed
On April 2, 2026, OpenAI replaced Codex's per-message pricing with token-based credit billing across all Plus, Pro, and Business plans. This was the biggest Codex pricing change of 2026 and makes older pricing guides misleading. The old model: one message equals roughly one fixed cost. The new model: cost tracks directly with tokens consumed — input tokens, cached input tokens, and output tokens at different credit rates.
The practical impact: a simple bug fix and a large codebase refactor now look very different on your bill even though they are both "one Codex task." A one-line fix sits at approximately 5 credits. A multi-file refactor that reads 30 files and generates 10 modified files can hit 45 credits — a 9x difference for what appears to be the same number of tasks. OpenAI itself says typical developer costs run $100-$200/developer/month for teams using Codex as a primary engineering tool.
All Six Plans — What You Actually Get
| Plan |
Price |
Cloud tasks / 5-hr window |
Included platforms |
Best for |
| Free |
$0 |
Very limited |
Web, CLI |
Evaluating, quick tasks only |
| Go |
$8/mo |
Low — no cloud task features |
Web, CLI, IDE |
Very light users. Most devs should skip to Plus |
| Plus (recommended) |
$20/mo |
15-80 tasks (GPT-5.5) |
Web, CLI, IDE, iOS |
Daily developers — best price/performance |
| Pro 5x |
$100/mo |
50-300 tasks |
All + Codex-Spark preview |
Full-time engineers, heavy agentic workflows |
| Pro 20x |
$200/mo |
200-1,200 tasks |
All + priority |
Power users, agencies, all-day Codex use |
| Enterprise |
Custom |
Shared credit pool (no fixed limits) |
All + SCIM, audit logs, RBAC |
Teams, compliance requirements |
API Pricing — Every Model and What It Actually Costs
| Model |
Input / 1M tokens |
Output / 1M tokens |
Best for |
| codex-mini-latest |
$1.50 |
$6.00 |
Routine tasks, CI/CD, bulk code generation |
| gpt-5.1-codex-mini (high) |
$0.25 |
$2.00 |
Simple refactors, test generation, boilerplate |
| gpt-5.3-codex (xhigh) |
$1.75 |
$14.00 |
Complex architecture, multi-file agentic tasks |
| gpt-5.5 (via subscription) |
$5.00 (API equiv) |
— |
Subscription only — 125 credits per 1M input tokens |
API prices are for direct API access without ChatGPT subscription. Regional data-residency endpoints carry a 10% uplift. API users get new models later than ChatGPT subscribers.
Subscription vs API — When Each Makes Sense
Subscription (Plus/Pro) makes sense when:
You are a developer using Codex interactively. At $20/month Plus, running 200 cloud tasks via the API at gpt-5.5 rates would cost approximately $50-$90 — 2.5-4.5x the subscription price. The flat rate subsidises moderate to heavy interactive use. Subscriptions also get new models faster and include cloud features (GitHub code review, Slack integration) that API-only access does not.
API makes sense when:
You are building a product that calls Codex programmatically. You need Codex in CI/CD pipelines or automation without a ChatGPT subscription. You want fine-grained control over model selection and cost per task. Note: API access excludes GitHub code review and Slack integration, and newer models are available later than to subscription users.
Codex Pro 5x ($100) vs Claude Code Max 5x ($100) — The Real Comparison
OpenAI introduced the Pro 5x tier on April 9, 2026 specifically to create a $100 option competing with Claude Code's $100 Max 5x. The comparison that actually matters:
| Dimension |
Codex Pro 5x ($100) |
Claude Code Max 5x ($100) |
| Paradigm |
Async — queue tasks, review PRs |
Terminal session — interactive |
| SWE-bench score |
85.5% autonomous (GPT-5.5) |
~69.2% (Opus 4.8) |
| Windows computer use |
Yes — GUI desktop automation |
No |
| Mobile coding |
Yes — iOS + Android app |
No |
| MCP integrations |
Limited |
3,000+ |
| CI/CD integration |
Native GitHub — async PRs |
GitHub Actions — interactive |
| Verdict |
Better benchmark, async workflow |
Better MCP ecosystem, interactive |
The most honest framing: Codex Pro 5x and Claude Code Max 5x are not the same product at the same price. Codex is an async agent you delegate to; Claude Code is a terminal session you work with. Most teams that buy both find they use them for different tasks — Codex for queued engineering work, Claude Code for complex interactive sessions. See our full Codex vs Cursor comparison for the daily use case breakdown.
Frequently Asked Questions
Is Codex included in ChatGPT Plus?
Yes. Codex is included in every ChatGPT plan from Free upward, including Plus ($20/month). Plus gives you Codex on the web, CLI, IDE extension, and iOS with a 5-hour rolling window of approximately 15-80 GPT-5.5 cloud tasks depending on task size. There is no separate Codex subscription — you pay for ChatGPT and Codex is bundled.
How much does Codex cost for a typical developer per month?
OpenAI itself estimates $100-$200/developer/month for teams using Codex as a primary engineering teammate. For individual developers on Plus ($20/month) doing moderate use, the flat subscription is significantly cheaper than equivalent API billing — 200 cloud tasks via API at gpt-5.5 rates would cost $50-$90. Most individual developers on Plus never exhaust their window.
What is the difference between Codex Pro 5x and Pro 20x?
Pro 5x at $100/month (added April 9, 2026) gives 5x the usage of Plus — roughly 50-300 cloud tasks per 5-hour window. Pro 20x at $200/month gives 20x Plus limits — 200-1,200 tasks per window — plus the original Pro features. Pro 5x was introduced to compete directly with Claude Code's $100 Max tier. For most active individual developers, Pro 5x is sufficient. Pro 20x is for power users and agency-level usage.
Related: Codex vs Cursor · Claude Code pricing · Four-way coding tool comparison · Best AI coding agents 2026