Claude Code vs Codex: plans, limits and features compared

Claude Code vs Codex comes down to how you pay and where you work. Both are coding agents that read a codebase, edit files and run commands. Claude Code needs a paid Claude plan from $20 a month. Codex is included in ChatGPT's free tier and every paid plan, from $8 a month. Prices are as of October 27, 2026.

This comparison is based on the products' docs and published results. We did not run a hands-on test, so it says nothing about which one writes better code on your project.

Claude Code vs Codex at a glance

Sources: Claude pricing and Codex pricing, read October 27, 2026.

What each plan costs

Claude Code is not in the free Claude plan. It comes with Pro ($20 a month, or $17 a month billed annually), Max (from $100 a month, with five times or twenty times Pro's usage per session), Team and Enterprise. Team Standard seats cost $25 a month, or $20 billed annually, and Team Premium seats cost $125, or $100 billed annually. Chat and Claude Code draw from the same usage pool. When you hit a limit you can wait, upgrade, or turn on usage credits billed at API rates. Claude's pricing page lists the Max tiers as "From $100" and gives no separate price for the 20x tier.

Codex is in every ChatGPT plan: Free ($0), Go ($8 a month), Plus ($20 a month), Pro (plans at $100, $200 or $500 a month), Business and Enterprise. Business is $25 per user a month, or $20 billed annually. Enterprise and Edu are priced through sales.

At $20 a month the two are level on sticker price. The gap is at the bottom, where Codex works on a free account and Claude Code does not, and at the top, where both start heavy-use plans at $100.

How the usage limits differ

Claude resets limits on a rolling five-hour window and adds weekly caps on paid plans. It does not publish a message count. Max plans are described only as multiples of Pro.

OpenAI publishes estimates. As of October 27, 2026, Plus and standard Business users get these local messages per five hours: 5 to 45 on GPT-6 Astra, 15 to 160 on GPT-6.1 Sol, 35 to 285 on GPT-6.1 Terra and 350 to 3,000 on GPT-6 Luna. OpenAI says Pro currently has no five-hour limit, and that weekly limits may also apply. The wide ranges show that the cost of a task depends heavily on its size and the model you pick.

Metered use is priced differently too. Claude's API charges per million tokens, for example $2 input and $10 output for Sonnet 5.5 and $4 and $20 for Opus 5.5. Codex on a credit plan charges credits per million tokens, for example 50 credits input and 250 output for GPT-6.1 Sol. OpenAI's page says API token prices are separate from subscription usage and should not be used to estimate included tasks. For how to see what you have used, see our guides to checking Codex token usage and checking Claude token usage.

Where you can use each one

Claude Code's docs list the terminal, VS Code, JetBrains, a desktop app and the web at claude.ai/code, with the mobile app as another way in. All of them share one engine, so CLAUDE.md files, settings and MCP servers carry across. It also runs in GitHub Actions and GitLab CI/CD, and the Agent SDK lets you build your own agents on top of it.

OpenAI's docs list the Codex CLI, an IDE extension and Codex Cloud, plus the ChatGPT desktop, mobile and web apps. The docs have sections on sandboxing, permission profiles and auto-review. The CLI installs with curl -fsSL https://chatgpt.com/codex/install.sh | sh, with Windows, npm and Homebrew options as well.

What the benchmarks show

We could not read agent-level scores for either product this week, so we are not quoting any. On October 27, 2026 the Terminal-Bench 4.0 leaderboard at tbench.ai loaded with column headers but no rows, and the pages for SWE-bench and Artificial Analysis's Coding Agent Index v2.0 did not expose scores for Claude Code or Codex in text we could read.

What Artificial Analysis does show is model-level, from its Capability Index comparison, with no date on the table: Claude Opus 5.5 ranked first (60), Claude Sonnet 5.5 second (58), Claude Fable 5.1 third (56), GPT-6 Astra sixth (55) and GPT-6.1 Sol seventh (54). That compares models, not the agents wrapped around them, and the gaps are small. Check the Terminal-Bench, SWE-bench and Artificial Analysis pages before you decide.

Which one fits

Pick Codex if you want to start free, already pay for ChatGPT, or want published message estimates. Pick Claude Code if you want one agent across terminal, IDE, web and CI with the same project files, or if you already use Claude for chat. For how the two companies' plans stack up beyond coding, see our ChatGPT vs Claude plans and benchmarks comparison.

Both have a free way to try the workflow before paying: Codex on the Free plan, and Claude Code through a Pro month you can cancel. Run the same task in each on your own repository, then compare the diffs.