Claude Code, Codex or Cursor: which fits how you work?

Guide · checked against the docs on Oct 15, 2026

All three of these tools can read your project, edit files and run commands for you. Where they differ is in the plan you pay for, the models you get with it, and where you'll spend your day: a terminal, an editor or a browser tab. We aren't going to pick a winner. Everything below comes from each vendor's own pages, read on October 15, 2026, so you can line them up and decide.

Not affiliated with or endorsed by Anthropic, OpenAI or Cursor. Plans and prices change often. Check each vendor's page before you pay. Prices exclude tax.

Our tool pages: Claude Code · OpenAI Codex · Cursor

At a glance

Price and plans

Claude Code. Anthropic's pricing page lists Claude Code on every paid plan and not on Free. Pro is $20 a month, or $17 a month if you pay $200 for the year. Max starts at $100 a month and gives 5x or 20x Pro's usage. Max also includes $100 to $200 a month in API credits. Team and Enterprise plans include Claude Code too, and Enterprise is listed at $20 per seat plus usage at API rates. The page also says that for heavy coding you can switch to pay-as-you-go API credits through a Console account.

Codex. OpenAI's Codex pricing page says Codex is included with ChatGPT Free, Go, Plus, Pro, Business, Edu and Enterprise. Free is $0, Go $8, Plus $20, and Pro comes in $100, $200 and $500 a month versions. Business is listed at $25 per user per month when billed monthly. You can also run Codex with an API key and pay API prices, which the page pitches for automation like CI.

Cursor. Cursor's pricing page now says plans are sold through X Subscription Corp and bundle Grok and X features (a checkmark, fewer ads and so on) along with Cursor. Free is $0 with limited requests. Plus is $8 a month, and the page says "Only Grok models are available on this plan". Premium is $30, Super $100 and Ultra $200, with "4x", "15x" and "50x" the usage of Plus. Teams starts at $30 per user per month. On Premium and up, models from other vendors are "billed separately by usage" at API list price.

Which models each can use

• Claude Code uses Anthropic's Claude models. The pricing page lists Haiku and Sonnet on Free, and adds Opus on Pro and up. Its Fable model is available through usage credits on Pro and capped at "50% of weekly limits" on Max. The Claude Code docs say the terminal CLI, VS Code and JetBrains versions also support third-party providers (Amazon Bedrock, Google Cloud's Agent Platform and Microsoft Foundry). That support covers those three, not Claude Code on the web: cloud sessions need a Claude subscription. The desktop app is a partial exception, because it can be routed through a gateway or a separate third-party desktop setup, per the feature availability page.
• Codex uses OpenAI's models. The pricing page lists GPT-6 Luna on Free and Go (in the desktop app, "subject to rollout"), and GPT-6.1 Sol and GPT-6 Luna on Plus. With an API key, "model availability follows the API models available to your key".
• Cursor lists models from several vendors. Its docs model table includes models from Anthropic, OpenAI, Google, Moonshot, Z.ai, Meta and Cursor itself, plus Grok models listed as "jointly trained by Cursor and SpaceXAI". On Plus, only Grok models are available. Other vendors' models are billed separately at API prices, per the pricing page. The pricing page lists third-party models at API list price on Premium and up. The docs' model table may be out of date compared with the pricing page, so go by the pricing page for what your plan includes.

Where it runs

• Claude Code runs in a terminal, in VS Code and JetBrains, in a desktop app, and on the web at claude.ai/code, according to its overview docs. The web version is meant for long tasks you check on later. Routines run in the cloud "even when your computer is off". The same docs cover GitHub Actions, GitLab CI/CD and Slack.
• Codex runs on the web, in a CLI, in an IDE extension, on iOS and in a ChatGPT desktop app, per its pricing page. Codex cloud, GitHub pull request reviews and the Slack integration need a paid ChatGPT plan, and none of them work with an API key.
• Cursor is first of all an editor. It also has a CLI that runs agents in the terminal, in scripts and in GitHub Actions. Cloud agents and automations are on Premium and up, per the pricing page.

Data and training settings

These are the defaults each vendor's own page states. Check your account settings, since they can change.

• Claude: the pricing page lists model training as "Opt-out" on Free, Pro and Max, and "None by default" on Team and Enterprise.
• Codex: the pricing page lists "No training on API or business data by default" for Business, Enterprise/Edu and API key use. It doesn't make that claim for Plus or Pro.
• Cursor: the pricing page says that with Privacy Mode on, "code data is not used for training by us or our model providers". Team-wide privacy settings and custom data retention are on Teams and Enterprise. The docs note that some Anthropic models need data retention approval when Privacy Mode is on.

Limits

None of the three gives a fixed message count. What they do say:

• Claude: limits reset on a rolling five-hour window, and paid plans add weekly limits. Chat and Claude Code share one pool. Pro gives "at least 5x" Free's usage per session.
• Codex: usage depends on the model and the task. The page gives rough ranges of local messages per five hours on Plus, for example "15-160" for GPT-6.1 Sol. Pro plans "currently have no five-hour limit". Weekly limits may apply, and you can buy credits when you run out.
• Cursor: each tier gets a pool of usage shared between Cursor and Grok, and it "refreshes weekly". When it runs out you can buy top-offs or wait.

Which fits you

There's no right answer here, only a better match for how you already work.

• You live in the terminal. All three have a CLI now. Claude Code's docs describe its CLI as "full-featured". Codex comes with a ChatGPT plan you may already pay for, and Cursor has a CLI too. On Plus, only Grok models are available. Other vendors' models are billed separately at API prices, per the pricing page. Look at which plan you already have, and what it includes.
• You live in an editor. Cursor is the editor, with the model menu built in. Claude Code has VS Code and JetBrains extensions, and Codex has an IDE extension, so you may be able to keep the editor you have. If you want models from other vendors in Cursor, note that no plan includes them: on Premium and up they're billed separately at API list price, and Plus is Grok only.
• You hand off tasks and check back later. Claude Code on the web and its routines, Codex cloud, and Cursor's cloud agents all do this. Each needs a paid plan, and for Cursor it's Premium or higher. If you work through GitHub, look at which ones review pull requests on the plan you'd buy.
• Your code is private or under license. Read the training defaults above for the exact plan you'd use. On some plans, keeping your code out of training is something you have to switch on yourself.

The cheapest way to decide is to try each one on a small real task from your own project, on its free or lowest plan, and keep notes. Our guide to choosing an AI tool for FiveM has a test checklist you can reuse.

We haven't published hands-on test results for these tools yet. This page compares what the vendors say, not how the tools performed for us.

Not affiliated with or endorsed by Anthropic, OpenAI or Cursor. Plans and prices change often; check each vendor's page before you pay.