OpenAI Codex vs Claude Code Pricing 2026: Both Bundled, Both Token-Metered
Neither OpenAI Codex nor Claude Code is sold as a standalone product. Both ride on a parent subscription — Codex inside ChatGPT, Claude Code inside Anthropic’s Claude plans — and both moved to token-based metering in 2026. The tiers even line up almost exactly: $20 entry, $100 mid, $200 top. That symmetry makes the real comparison less about price and more about which ecosystem and usage model fits how you code.
The Plans Side by Side
| Tier | OpenAI Codex (via ChatGPT) | Claude Code (via Claude) |
|---|---|---|
| Free | ChatGPT Free — Codex included, limited, strict rate caps | None — no free Claude Code tier |
| Entry | Plus — $20/mo | Claude Pro — $20/mo |
| Mid | — | Max 5x — $100/mo (5x Pro usage) |
| High | Pro — $200/mo (minimal rate limits) | Max 20x — $200/mo (20x Pro usage) |
| Team | Business — per seat | Team — $30/seat/mo |
| Metering | Token-based credits (since Apr 2, 2026) | Token-based usage limits / API pay-per-token |
The Free Tier: Codex Wins by Default
This is the cleanest difference. ChatGPT Free includes Codex — limited, with strict rate caps, but enough to try it without paying. Claude Code has no free tier at all; the cheapest way in is Claude Pro at $20/month.
If “can I try it for $0” matters, Codex is the only option here.
Entry: ChatGPT Plus ($20) vs Claude Pro ($20)
Same price, both bundle the coding agent.
ChatGPT Plus — $20/month includes Codex with standard rate limits. Since OpenAI moved Codex to token-based billing on April 2, 2026, your subscription provides a monthly credit allocation, and each task consumes credits based on model and usage. A typical Plus user runs ~$20–40/month effective once light overage is included.
Claude Pro — $20/month includes Claude Code with standard usage limits. The constraint is the same token reality: Claude Code reads multiple files, plans, and iterates, so a single complex refactor can eat a large chunk of your daily Pro allowance. The 5x/20x multipliers above Pro apply across all Claude usage — chat and Projects included — so heavy non-coding Claude use eats into your coding headroom.
At $20, the choice comes down to which model you trust for code. Both meter by tokens; both will throttle a heavy day. There’s no price advantage either way at entry.
Mid and High: $100 and $200 Tiers
Here the ladders differ in shape.
Claude Code has a clean three-step consumer ladder: Pro $20 → Max 5x $100 (five times Pro usage, the practical daily-driver tier) → Max 20x $200 (twenty times Pro, for large refactors, codebase-wide changes, and migrations where limits rarely bite).
Codex jumps from Plus $20 straight to Pro $200 (minimal rate limits, for full-time Codex + CLI users), with no clean $100 middle rung — heavy Plus users instead pay overage, landing around $100–150/month effective before Pro becomes worth it.
So if your usage sits in the middle — more than $20 of activity but not full-time — Claude’s Max 5x at $100 is a defined tier, whereas Codex makes you choose between paying overage on Plus or jumping to the $200 Pro plan. For predictable mid-level budgeting, Claude’s ladder is friendlier.
Teams and API
Claude offers Team at $30/seat/month and an API pay-per-token option with no fixed limits — useful if you want to wire Claude Code into automated pipelines without a consumer cap. Codex Business is per-seat with team management and admin, billed through ChatGPT’s business tier. Both support direct token billing for automation; pick by which model your team standardizes on.
Which Should You Pick?
- Want to try for free? Codex via ChatGPT Free — Claude Code has no free tier.
- Light, occasional coding at $20? Either Plus or Claude Pro; choose by preferred model.
- Daily driver, mid-level usage? Claude Max 5x ($100) gives a clean tier; Codex makes you pick overage or the $200 Pro.
- Full-time, heavy usage? Codex Pro ($200) or Claude Max 20x ($200) — both remove most limits.
- Need pipeline automation? Claude’s API pay-per-token is the most flexible.
For full breakdowns, see OpenAI Codex pricing and Claude Code pricing. If you also use a dedicated IDE, Cursor pricing and ChatGPT pricing round out the picture.
The Bottom Line
The tiers are nearly mirror images — $20, then $200 at the top — and both meter by tokens, so neither is structurally cheaper. The deciding factors are: Codex has a free tier and Claude doesn’t; Claude has a clean $100 mid-tier and Codex doesn’t. Try Codex free, but if you’re a daily-driver who wants predictable mid-level cost, Claude’s Max 5x is the better-shaped plan.
Compare every AI coding tool’s pricing side by side → Best AI app builders pricing 2026