Codex vs Claude Code: Pricing, Usage Limits and How They Differ
Already paying for ChatGPT? Start with Codex. Paying for Claude? Start with Claude Code. Want to try free? Only Codex offers that. Compares October 2026 pricing (both from $20, Codex also on Free and Go), usage limits, workflows, China availability and when paying for both makes sense.
If you already pay for ChatGPT, start with Codex. If you already pay for Claude, start with Claude Code. If you pay for neither and want to try a coding agent for free, Codex is the only one of the two with a free entry point. Both are agents that edit code and run tests on their own, and day to day they do similar work; what separates them is how you pay and how usage is counted. Codex comes with every ChatGPT plan, including the free plan and the $8 Go plan through the desktop app, while Claude Code starts with Claude Pro at $20 a month. Neither is offered in mainland China.
Yes, Free and Go include the desktop app (rolling out)
No, Claude Free does not include Claude Code
Full feature set from
ChatGPT Plus, $20/month
Claude Pro, $20/month ($17/month billed annually)
Heavy-use plans
Pro at $100, $200 or $500/month, currently no five-hour limit
Max from $100/month at 5x or 20x Pro usage
Where it runs
Desktop app, web, CLI, IDE extension, mobile, cloud tasks
Terminal, VS Code and JetBrains extensions, web, desktop app, mobile, Slack
How usage is counted
Plus lists per-model message ranges per five hours, with possible weekly limits
One pool shared with Claude chat, with five-hour and weekly limits
Open source
The Codex CLI is open source under Apache-2.0
Not open source
Mainland China
Not an OpenAI-supported region
Not an Anthropic-supported region
How Codex and Claude Code pricing works
Both start at $20 a month for the full feature set. The differences sit below and above that line.
Below $20, Codex has an entry point Claude Code lacks. ChatGPT Free and the $8 Go plan both include the Codex desktop app running GPT-6 Luna at standard speed. If you want to see what working with a coding agent feels like before spending anything, Codex is your only option here. OpenAI does not publish limits for Free and Go, and the web app, CLI and IDE extension start at Plus.
Above $20, the two take different approaches. ChatGPT Pro comes in $100, $200 and $500 tiers; OpenAI says Pro currently has no five-hour limit, and the $500 tier adds an Ultrafast mode. Claude Max comes in two sizes from $100 a month, 5x and 20x the Pro allowance, and still carries both five-hour and weekly limits.
One detail is easy to miss. Claude Code shares its allowance with your Claude chats, so a heavy day of chatting leaves less for coding that evening. Codex on Plus lists estimated local-message ranges per five hours for each model, for example 15 to 160 messages with GPT-6.1 Sol and 350 to 3,000 with GPT-6 Luna, and notes that the real number depends on task size.
Both also accept an API key with pay-as-you-go billing, which suits people whose usage swings too much for a subscription.
How they differ in daily use
The two agents can do much the same work. Both read a codebase, edit several files, run tests and open pull requests, and both can be set to ask before editing files or running commands. The differences are where you use them and how they fit your workflow.
Codex leans toward its desktop app and the cloud. It can work on local code from the ChatGPT desktop app, hand tasks to Codex Cloud to run remotely, and let you check or start tasks from your phone. The Codex CLI is open source, so you can read exactly how it calls models and what it changes.
Claude Code leans toward the terminal and the editor you already use. It runs locally in your terminal and talks directly to model APIs, with no extra backend. Anthropic ships VS Code and JetBrains extensions, and the VS Code extension also works in Cursor. You can start tasks from Slack and run it in GitHub Actions. On Pro, Max and Team plans, auto mode is on by default, letting it work longer while still catching risky commands.
Picking by habit. If you like managing several tasks in a dedicated app and often push work to the cloud, Codex fits that style. If you spend the day in a terminal and an editor and want the agent working in the next pane, Claude Code feels more natural.
Can you use Codex or Claude Code in China?
Neither is officially available.
Codex uses your ChatGPT account, so OpenAI's regional rules apply. OpenAI's supported countries list excludes mainland China, Hong Kong and Macau, and warns that use outside listed locations may get an account blocked or suspended.
Claude Code follows Anthropic's rules. Anthropic's supported countries list also leaves out mainland China, Hong Kong and Macau, and that policy covers Claude.ai, Claude Code and the API. Taiwan is on both lists.
Neither company's pages mention Alipay or WeChat Pay. Reports from developers in China suggest both need a proxy for a stable connection there, that dual-currency cards issued in mainland China are often declined at web checkout, and that subscribing through an overseas App Store or Google Play account is the more common route. Those communities also report accounts being deactivated, usually linked to frequent IP changes, virtual cards or resold subscriptions. This is community experience rather than official policy, and it can change at any time.
Developers in mainland China who want a coding agent without these issues can look at domestic alternatives, or use Cursor with the models it allows in their region; our Claude Code vs Cursor comparison covers that route.
Is paying for both worth it?
The cheapest way to have both is ChatGPT Plus plus Claude Pro, $40 a month. Whether that makes sense depends on whether you also want both companies' chat assistants.
If you already chat in both ChatGPT and Claude, keeping both subscriptions effectively gives you a second coding agent at no extra cost. You can switch when one allowance runs out, or give the same task to both and keep the cleaner change.
If you are paying only to write code, pick one and use it hard for a month. Note how many days a week you hit the limit and how long it blocked you. If it rarely happened, a second subscription is unnecessary. If it happened often, compare the cost of a higher tier from the same company before adding the other one.
FAQ
Is Codex or Claude Code better?
They can do much the same work, so the choice mostly comes down to which subscription you already pay for and where you like to work. Choose Codex if you prefer a dedicated app and cloud tasks, and Claude Code if you live in the terminal and your editor.
How does Codex vs Claude Code pricing compare?
Both start at $20 a month for the full feature set. Codex is also available through the desktop app on ChatGPT Free and the $8 Go plan, while Claude Code has no free entry. Heavy-use plans start at $100 a month on both sides.
Is Codex free?
ChatGPT Free includes the Codex desktop app as it rolls out, without published limits. The web app, CLI and IDE extension require Plus or higher.
Can I use Codex and Claude Code together?
Yes. Both can run in the same repository, with separate accounts and separate limits. A common pattern is to switch when one runs out, or to let each produce a version and keep the better one.
