OpenAI Codex Review 2026: Pricing, Features and Verdict

What it is

Codex is OpenAI's coding agent, sold as part of ChatGPT. It works on your code locally (CLI, IDE extension, desktop app) or in OpenAI's cloud, where it can take a task, run it in a sandbox and come back with changes. On Plus and up it also does automatic code review on GitHub and works from Slack.

Codex and ChatGPT's agentic work features share one usage allowance, so heavy Codex use and heavy ChatGPT agent use draw from the same pool.

Pricing explained

Plan	Price	Codex access
Free	$0	GPT 6 Luna at Standard speed in the desktop app, subject to rollout
Go	$8 a month	Same as Free
Plus	$20 a month	Web, CLI, IDE extension, iOS, code review, Slack, GPT 6.1 Sol and GPT 6 Luna
Pro	$100, $200 or $500 a month	Everything in Plus, no 5 hour limit currently, Ultrafast mode on $500
Business	$25 per user monthly, $20 annually (2+ users)	Larger cloud VMs, SAML SSO, no training on business data by default
Enterprise and Edu	Custom	SCIM, EKM, RBAC, audit logs, data residency
API key	API pricing	CLI, SDK and IDE extension only, no cloud features

Prices as listed on OpenAI's Codex pricing page in October 2026.

Plus usage, in numbers. OpenAI publishes estimates of local messages per 5 hour period on Plus and Standard Business:

Model	Plus and Standard Business, per 5 hours
GPT 6 Astra	5 to 45
GPT 6.1 Sol	15 to 160
GPT 6 Sol	15 to 150
GPT 6 Luna	350 to 3,000

These are estimates, not fixed limits, and weekly limits may also apply. Cloud tasks can use more than local messages. Fast mode uses included limits at 2.5x and Ultrafast at 8x.

What $20 buys. Plus covers ChatGPT and Codex. If you already pay for ChatGPT, Codex costs nothing extra; if you do not, it is priced the same as Claude Pro for Claude Code.

Overflow options. Plus and Pro users can buy ChatGPT credits instead of upgrading. Any user can run extra local sessions on an API key at API rates.

Team math. Five users on Business cost $125 a month, or $100 billed annually.

See the category in the Price Index.

Features

Many surfaces. CLI, IDE extension, desktop app, web and iOS on Plus and up. The desktop app also supports voice, billed against your usage budget at $0.05 per minute.

Cloud tasks and review. Codex can run work in the cloud and do automatic code review on GitHub, plus a Slack integration, on Plus and up. Business gets larger virtual machines for cloud runs.

Model lineup. GPT 6.1 Sol for most coding, GPT 6 Astra for the heaviest tasks (and the lowest message count), GPT 6 Luna for fast, cheap work. GPT 5.5 retired from Codex with ChatGPT sign in on October 14, 2026.

API key mode. For CI and shared environments, Codex runs on an API key in the CLI, SDK or IDE extension. Cloud features are not available in that mode.

Limits behavior. If you hit the limit mid turn, the agent can finish that turn, subject to fair use, which avoids half applied edits.

Image generation. Codex can generate images, but OpenAI says image turns use included limits about 3 to 5 times faster on average than similar turns without images, and image generation is not available on the Free plan. If you use it for UI mockups, expect your 5 hour allowance to shrink accordingly.

Setup

Sign in with your ChatGPT account in the desktop app, CLI or IDE extension. In the CLI, /status shows remaining limits during a session, and a usage dashboard shows current limits and reset times. Business workspaces add SSO and admin controls; Enterprise adds the Compliance API for audit logs.

Who it is for

• ChatGPT Plus and Pro subscribers, for whom Codex is already paid for.
• Developers who like to delegate tasks to cloud sandboxes and review the results.
• Teams that want GitHub code review and Slack inside one subscription.
• Very heavy users, since Pro currently has no 5 hour limit.

Who should skip it

• Free users who want a real agent; Free and Go access is desktop only and subject to rollout.
• Developers who want Claude or Gemini on the same plan; GitHub Copilot includes several labs' models for $10.
• Anyone who needs predictable throughput on Plus; 15 to 160 messages per 5 hours on GPT 6.1 Sol is a wide range.
• JetBrains users, who should confirm IDE support for their editor before buying, since the pricing page lists an IDE extension without naming editors.

Verdict

Codex is a strong agent with the widest set of surfaces after Copilot (CLI, IDE, desktop, web and iOS) and the most generous top tier, since Pro plans currently have no 5 hour limit. It ranks third in our best AI coding agents list behind GitHub Copilot, which is cheaper and bundles more labs' models, and Claude Code, which has the stronger extension rating (3.7 from 829 against Codex's 2.9 from 721). If you already pay for ChatGPT, start here before buying anything else.