Codex vs Claude Code (2026): Prices, Limits and Which One to Pick

Short answer (November 2026): Codex and Claude Code both start at $20 a month on their main paid plans. Pick Codex if you want to try a coding agent for free, or if you want strict safety limits switched on from the start. Pick Claude Code if you already use Claude, or if you want more ways to customize how the agent works. Both can edit real projects and run commands, so the best test is your own code.
We checked plans and defaults on OpenAI's and Anthropic's own pages in November 2026. Plans change often, so check Codex pricing and Claude pricing before you pay.
Codex vs Claude Code: Plans and Prices
| Codex | Claude Code | |
|---|---|---|
| Free option | Yes, on ChatGPT Free, desktop app only, subject to rollout | No, Claude Free does not include it |
| Budget plan | ChatGPT Go, $8/month (same limited access as Free) | None |
| Main paid plan | ChatGPT Plus, $20/month | Claude Pro, $20/month ($17/month billed yearly) |
| Power-user plans | ChatGPT Pro: $100, $200 or $500/month | Max: $100/month. Ultra: $200/month |
| Pay by usage | API key, or buy ChatGPT credits | API account, or usage credits on paid plans |
| Models | OpenAI models, such as GPT-6.1 Sol | Anthropic models: Opus, Sonnet and Haiku, plus Fable on some plans |
Prices are US prices. Some countries get local pricing.

Codex is included from the Free plan up, with limited access on Free and Go. Source: OpenAI Codex pricing, captured November 9, 2026.

Claude Code is included from Pro up. Source: Claude pricing, captured November 9, 2026.
If you want more detail on Claude's plans, see Claude Code on Pro vs Max and Claude Free vs Pro vs Max.
Usage Limits
Both tools have limits on the main plans, and both depend on how heavy your work is.
Codex on ChatGPT Plus: OpenAI publishes estimates per 5 hours. It lists about 15 to 160 local messages on GPT-6.1 Sol and 350 to 3,000 on GPT-6 Luna, its lighter model. Weekly limits may also apply, and cloud tasks can use more of your allowance. ChatGPT Pro currently has no 5-hour limit.
Claude Code on Pro: Anthropic does not publish a message count. Your allowance resets every 5 hours and every week, and it runs out faster with bigger files, longer sessions and stronger models. Max gives 5 times the Pro usage. Ultra gives 10 times the Pro usage each week and has no 5-hour limit for verified accounts.
Because one company counts messages and the other doesn't, you can't compare the two allowances directly. If you code for hours every day, expect to need more than a $20 plan on either side.
Safety Defaults: The Biggest Practical Difference
A coding agent can change files and run commands on your computer, so its default limits matter, especially when you are learning.
Codex runs inside an operating system sandbox by default. In normal local use it can edit files in your project, but it has no network access unless you allow it. It asks before going outside those limits.
Claude Code asks your permission before actions like running commands. Its own sandbox is off by default. You can turn it on with /sandbox. It works on macOS, Linux and WSL2, but not on native Windows.
For a beginner, Codex's default is the safer starting point. On Claude Code, read each request before you approve it, and consider turning the sandbox on.
Either way, use Git and commit before big changes. Then you can undo anything the agent gets wrong.
Watch What You Approve Forever in Claude Code
When Claude Code asks to run a command, you can pick "Yes, and don't ask again". Claude Code then saves that command as an allow rule in .claude/settings.local.json. If you typed a password or API key into the command, the key is saved too.
In November 2026, security firm Apiiro reported finding real secrets in 1,122 public copies of this file on GitHub.
Current Claude Code docs say it adds the file to your global git excludes the first time it writes it. So the risk is mostly older projects and files made by hand.
To stay safe:
- Keep keys in environment variables or a secret manager.
- Never type a key into a command.
- Search your public repos for
.claude/settings.local.json.
Where Each One Runs
| Codex | Claude Code | |
|---|---|---|
| Terminal | Yes | Yes |
| Code editors | IDE extension | VS Code, Cursor, JetBrains |
| Desktop app | Yes, in the ChatGPT app | Yes, in the Claude app (paid plans) |
| Web and phone | Yes | Yes |
| Project instructions file | AGENTS.md | CLAUDE.md (can also read AGENTS.md) |
| MCP, skills, hooks, subagents | Yes | Yes |
| Open source | CLI is open source (Apache-2.0) | No |
The feature lists are now very close. Claude Code's docs go further on scheduled tasks and on moving a session between your phone, browser and terminal. Codex's main advantages are its open source CLI, its free starting point and its sandbox default.
Which One Is Better for Coding?
We don't rank them with benchmark scores. Results change with each new model, and published comparisons often disagree. Many developers keep both and switch when one gets stuck.
Try this test: give both tools the same small bug in a copy of your project. Compare whether the fix works, how much you had to explain, and how much of your limit it used.
Which One to Pick
Students and beginners
Start with Codex on the free plan if it is available to you. You can learn how a coding agent works without paying. When you are ready to pay, $20 on either side is enough for study projects.
You already pay for ChatGPT or Claude
Use the one included in your plan. Both come with the $20 plan, so try it before you buy anything else.
Heavy daily coding
Compare ChatGPT Pro and Claude Max or Ultra, starting at $100 a month. Try the $20 plan for a week first, and only upgrade if you keep hitting limits.
Teams with strict security rules
Start with Codex's default sandbox, or turn on Claude Code's sandbox and manage it through settings. Ask your security team before connecting either tool to company code.
Choosing between an editor and an agent
If you prefer working inside an editor, read Cursor vs Claude Code. For other terminal tools, see Copilot CLI vs Claude Code.
Frequently Asked Questions
Is Codex better than Claude Code?
Neither is better at everything. Codex has a free starting point and a sandbox on by default. Claude Code has deep customization and works well if you already use Claude. Try both on the same task.
Is Codex free?
Codex is included in ChatGPT Free, but with limited access: the GPT-6 Luna model in the desktop app, subject to rollout. For the CLI and cloud tasks with stronger models, expect to need a paid plan.
Is Claude Code free?
No. Claude's free plan does not include Claude Code. It starts with Claude Pro at $20 a month, or $17 a month billed yearly.
Can I use Codex and Claude Code together?
Yes. They are separate tools and can work on the same Git project. Run one at a time on a change, and review each result. Claude Code can also read an AGENTS.md file, so both can share project instructions.
Which is safer for beginners?
Codex, by default, because it starts inside a sandbox with no network access. Claude Code asks for permission, but its sandbox is off until you turn it on.
Learn Both for Free
FreeAcademy has free courses with certificates for both tools:
- OpenAI Codex: From CLI to Cloud Agents: learn Codex step by step
- Claude Code: AI-Powered Coding: go from the terminal to production
- Claude Code Advanced: agents, hooks, MCP servers and automation
All courses are 100% free, including the certificate.
Last checked: November 9, 2026. Plans, models and limits change often. Check each provider's pricing page for the latest details.
Enjoyed this article?
Join The FreeAcademy Weekly
One practical AI email every Tuesday. New free courses, AI tips, and a short note from the founder.
Free forever. Unsubscribe anytime.
Related articles

Cursor vs Claude Code in 2026: IDE Agent vs Terminal Agent for Coding
Cursor vs Claude Code for coding in 2026: IDE agent vs terminal agent, current plans from $8 to $200, why Claude costs extra in Cursor, and which to pick.

Claude Code on Pro vs Max: Which Plan Do You Actually Need?
Compare Claude Code Pro vs Max plans — pricing, usage limits, extended thinking, and speed. Find out which Anthropic plan fits your coding workflow in 2026.

Copilot CLI vs Claude Code: Terminal AI Showdown 2026
Copilot CLI vs Claude Code compared head-to-head for 2026: features, pricing, agentic capabilities, speed, and which terminal AI coding tool wins for you.

