Short answer: Codex if you already pay OpenAI, Cursor if you want all three vendors' models, Claude Code if you live in the terminal
Verified August 3, 2026. TL;DR: Codex is bundled into every ChatGPT plan, including the $0 one, so it costs nothing extra if you already pay OpenAI. Cursor is the only one of the three that runs OpenAI, Anthropic and Google models in the same editor. Claude Code has no free tier and is the terminal-first option. Pick by what you already pay for.
| Codex | Cursor | Claude Code | |
|---|---|---|---|
| Form | Web, CLI, IDE extension, iOS | IDE, CLI, iOS, cloud agents | Terminal, IDE, desktop, web |
| Free tier | ChatGPT Free, $0 | Hobby, $0 | None |
| Entry paid tier | Go $8, Plus $20 | Pro, $20/mo | Pro, $20/mo, $17 annual |
| Top individual tier | Pro, from $100, 5x or 20x Plus | Ultra, $200, 20x Pro limits | Max, from $100 |
| Models | OpenAI only | OpenAI, Anthropic, Google, SpaceXAI, Composer | Claude only |
| What it meters | Five-hour window, weekly may apply | Dollars of model usage | Five-hour window plus weekly |
Cursor is the only one that runs the other two's models
Cursor's model list carries GPT-5.6 Sol, Terra and Luna next to Claude Opus 5, Sonnet 5 and Fable 5, plus Gemini and Cursor's own Composer 2.5 and Grok 4.5. Codex runs OpenAI models, and on an API key availability just follows your key. Claude Code runs Claude models. This is not three equivalent products: it is two vendor agents and one neutral editor that hosts both.
Two of them meter time, one meters dollars
Codex and Claude Code bill in windows. OpenAI's Plus limits are 10 to 100 GPT-5.6 Sol messages per five hours, 25 to 200 on Terra and 250 to 2,000 on Luna, with additional weekly limits that may apply. Claude's allowance resets on a rolling five-hour window plus a weekly one, shared with Claude chat and Cowork. Cursor bills dollars: Pro includes $20 of third-party model usage a month, then usage continues at API rates.
The free tiers are not comparable
Codex is listed at $0 on ChatGPT Free, then $8 on Go and $20 on Plus. Cursor's Hobby plan is $0 too. Claude Code has no free tier: Anthropic's pricing FAQ says it is "included in all paid plans", which starts at Pro, $20 monthly or $17 billed annually. If you want to test a coding agent without handing over a card, that rules Claude Code out and leaves two.
Budget the top tier, not the entry tier
Sticker prices understate what a working seat costs. Anthropic's own cost guidance puts enterprise Claude Code at roughly $13 per developer per active day and $150 to $250 per developer per month, staying under $30 per active day for 90% of users. Cursor's Ultra tier is $200 a month and Codex Pro starts at $100. Plan for the tier above the one you think you need.
How to choose, in one line each
Already paying for ChatGPT, start with Codex, because it costs nothing extra. Want one editor that runs all three vendors' models, buy Cursor and read Claude Code vs Cursor first. Want a terminal-first agent tied to Claude models, buy Claude Pro. For the wider field, see the Codex, Claude Code and Copilot breakdown, Gemini CLI vs Claude Code vs Codex, or Antigravity vs Cursor vs Claude Code.