AI Tools

Codex vs Claude Code: which AI coding agent should you pay for?

Published 2026-08-24 · Researched and price-checked on official sources

Codex vs Claude Code is now a pricing fork: OpenAI's coding agent starts at a $0 free tier with an $8/mo Go plan, while Anthropic's Claude Code only unlocks at Claude Pro, $17/mo billed annually ($200 up front) or $20 month-to-month. Both sell $100/mo tiers with 5x-20x more usage, and both offer pay-per-token API access.

Key facts

What matters

Codex starts free at $0, with an $8/mo Go plan; Claude Code requires Claude Pro at $17/mo annual ($200 up front) or $20 month-to-month.
Both vendors sell high-usage tiers from $100/mo: Codex Pro offers 5x-20x higher rate limits than Plus, Claude Max offers 5x-20x more usage than Pro.
On the API route, gpt-5.3-codex costs $1.75 per million input tokens and $14 output; Claude Sonnet 5 costs $2 input and $10 output.
Codex runs in the CLI, IDE extension, web, iOS, and cloud; Claude Code runs in Terminal, VS Code, JetBrains, desktop, and web, and supports third-party providers.
Every price in this comparison was verified on official OpenAI and Anthropic pricing pages on August 24, 2026.

What are Codex and Claude Code?

Codex is OpenAI's AI coding agent. It takes natural-language tasks and works through them in your terminal, IDE, or browser, signing in with a ChatGPT account. Anthropic's official pricing page groups it into every ChatGPT plan: Free, Go, Plus, Pro, Business, Edu, and Enterprise.

Claude Code is Anthropic's coding agent, a terminal-first tool that edits files, runs commands, and delegates work to subagents. Per Anthropic's overview docs, it ships in Terminal, VS Code, a desktop app, JetBrains, and the web. Most surfaces require a Claude subscription or an Anthropic Console account, and the CLI, VS Code, and JetBrains additionally support third-party providers.

Both tools target the same buyer: a developer or team paying monthly for an AI pair-programmer instead of, or on top of, a traditional IDE assistant.

Codex vs Claude Code pricing, side by side

The table below compares the plans listed on OpenAI's official Codex pricing page and Anthropic's plans page, both fetched on August 24, 2026. Neither vendor publishes a head-to-head, so this is the side-by-side you would otherwise have to assemble from two pricing pages.

Codex subscription tiers: Free, Go, Plus, Pro

OpenAI's Codex pricing page lists five consumer and business tiers. The Free plan costs $0 and is described as a way to explore Codex on quick coding tasks. Go costs $8/mo and targets lightweight coding tasks. Plus costs $20/mo and adds Codex on the web, in the CLI, in the IDE extension, and on iOS.

Pro starts at $100/mo and promises 5x or 20x higher rate limits than Plus. It also includes access to GPT-5.3-Codex-Spark, a research-preview fast Codex model, and the $200/mo tier adds unlimited ChatGPT Voice. Business costs $20 per user per month billed annually with at least two users, or $25 per user month-to-month.

The same official page confirms that pay-as-you-go exists: use Codex in the CLI, SDK, or IDE extension and pay only for the tokens it consumes, based on API pricing.

Claude Code subscription tiers: Pro and Max

Anthropic's plans page shows a Free tier for chat, then Pro at $17/mo with the annual subscription discount ($200 billed up front) or $20 if billed monthly. The Pro description lists Claude Code as included, alongside Claude Cowork, Claude Design, and Claude Science.

Max comes in two levels, Max 5x and Max 20x, both from $100/mo, offering 5x or 20x more usage than Pro, higher output limits, and early access to advanced Claude features. Team and Enterprise pricing go through a sales questionnaire rather than a public price list.

Claude Code can also run against the Anthropic Console API: the overview docs show login either through a subscription or an ANTHROPIC_API_KEY, with third-party providers supported on the CLI, VS Code, and JetBrains.

API pricing: what you actually pay per token

When you run either agent against the API, the bill is per token. OpenAI's API pricing page lists the Codex model gpt-5.3-codex at $1.75 per million input tokens, $0.175 for cached input, and $14.00 per million output tokens.

Anthropic's pricing page lists Claude Sonnet 5 at $2 per million input tokens and $10 output, with prompt-cache writes at $2.50 and reads at $0.20. Opus 5 runs $5 input and $25 output, and Haiku 4.5 runs $1 input and $5 output.

The arithmetic matters: gpt-5.3-codex is cheaper on input ($1.75 vs $2) but pricier on output ($14 vs $10) than Sonnet 5. Long-agent sessions that generate lots of code tilt toward Sonnet 5; prompt-heavy workloads with little output tilt toward gpt-5.3-codex.

Who should pick Codex, who should pick Claude Code

Pick Codex if you want a zero-cost trial: the $0 Free tier and $8 Go plan give you a working agent before you commit, and Plus at $20/mo matches the price of a Claude Pro subscription.
Pick Claude Code if you already pay for Claude Pro or Max: the agent is bundled into those plans, so the marginal cost is zero and the Sonnet 5 output price undercuts gpt-5.3-codex.
Pick by ecosystem: Codex signs in with ChatGPT and covers iOS and cloud runs; Claude Code covers JetBrains and third-party providers, which matters if you are locked out of one vendor's account policy.
Pick by volume: teams on the API route should model their input/output mix against the token prices above before choosing.

The honest catch: limits, prepay, and apples-to-oranges claims

Usage-limit numbers are not directly comparable. Codex Pro's 5x-20x is relative to Plus rate limits; Claude Max's 5x-20x is relative to Pro usage. Each vendor defines its own usage unit, so a 20x claim on one side does not equal a 20x claim on the other.

Watch the billing structure. Claude Pro at $17/mo means $200 up front for the year, not a monthly commitment you can cancel after one month at that rate. Codex Pro is listed as from $100/mo, and the 20x rate-limit tier with unlimited voice is the $200/mo level.

Finally, both pages change often. The prices in this article were read directly from OpenAI's and Anthropic's official pricing pages on August 24, 2026, and the only safe way to plan a purchase is to recheck both pages the week you buy.

FAQ

Is Codex free?

Yes. OpenAI's Codex pricing page lists a $0 Free tier for quick coding tasks. Paid plans start at Go for $8/mo and Plus for $20/mo.

Does Claude Code require a paid Claude subscription?

Yes for the standard route: Claude Code is included with Claude Pro ($17/mo billed annually or $20 month-to-month) and Max (from $100/mo). You can also pay per token through the Anthropic Console API.

Which is cheaper, Codex or Claude Code?

At the entry level Codex is cheaper: a $0 free tier and an $8 Go plan, while Claude Code starts at Claude Pro. On the API route, gpt-5.3-codex costs $1.75 per million input and $14 output tokens, versus Claude Sonnet 5 at $2 input and $10 output, so the cheaper option depends on your input-to-output mix.