Claude Code vs Codex in 2026: which to pay for, and when to use both

Disclosure: Enjoy is our product, and it runs both agents. We checked every fact below against Anthropic’s, OpenAI’s and Artificial Analysis’s own pages on October 27, 2026, and we say where you don’t need Enjoy at all.

Claude Code vs Codex, compared

Claude Code, from Anthropic, and Codex, from OpenAI, do the same job: you describe a change, and the agent reads your project, edits files and runs commands until it’s done. Each comes with its own lab’s plans, so choosing one mostly means choosing which plan to pay for.

Option	Best for	Price, checked October 27, 2026	Runs on	Models
Claude Code	The hardest work	Claude Pro $20 a month, or $17 billed yearly; Max 5x $100; Max 20x $200	Mac, Windows, Linux	Opus 5.5, Sonnet 5.5, Haiku 5.5; Fable 5.1 on Max
Codex	A lower cost per task, and a free start	Free and Go ($8) in the desktop app, with GPT-6 Luna; Plus $20; Pro $100, $200 or $500	Mac, Windows, Linux	GPT-6.1 Sol, GPT-6.1 Terra, GPT-6 Astra, GPT-6 Luna
Both, each in its own app	Two labs’ limits, and a second opinion	From $40 a month: Claude Pro and ChatGPT Plus	Mac, Windows, Linux	All of the above
Enjoy	Both in one conversation	Free, on top of your plans; Enjoy Cloud adds your phone, other computers and teammates (pricing)	Mac, Windows, Linux	Claude Code, Codex, Grok Build, Antigravity, Muse Code, OpenCode

Prices are each lab’s US list prices per month, before tax. Both agents can also run on an API key that bills per token instead of a plan.

How the two differ

	Claude Code	Codex
Coding Agent Index, best setup	61.7, with Opus 5.5 at max effort	54.7, with GPT-6 Astra at max effort
API cost and time per task on the index	$11.42 and 61 minutes, with Opus 5.5	$1.50 and 24 minutes, with GPT-6.1 Sol, which scores 54.1
Cheapest plan with the full agent	Claude Pro, $20 a month	ChatGPT Plus, $20 a month
On a free plan	No	In the desktop app, with GPT-6 Luna
Published limits	Multiples of Pro, without message counts	15 to 160 messages every five hours with GPT-6.1 Sol on Plus
Open source	No	The CLI, under Apache 2.0
On Windows	Native or WSL; sandboxing needs WSL 2	Native, with its own Windows sandbox
Project instructions	CLAUDE.md, or AGENTS.md when there’s none	AGENTS.md

The scores come from Artificial Analysis’s Coding Agent Index, which runs each agent and model together on 290 tasks from three benchmarks, as it stood on October 27. Its costs are “pay-per-token API cost, not consumer plan pricing”: on a plan, a task draws on your limits instead. Our memo on the index covers how it works.

Claude Code

Best for: the hardest work, when results matter more than cost.

Anthropic calls Claude Code “an agentic coding tool that reads your codebase, edits files, runs commands, and integrates with your development tools,” available “in your terminal, IDE, desktop app, and browser.” The terminal version runs on macOS, Windows and Linux, and the desktop app on Mac and Windows, with Linux in beta.

It holds the top two places on the index among setups that use one model: Opus 5.5 at max effort scores 61.7, and Sonnet 5.5 at max effort 59.3. And when a limit stops it partway through a task, its docs say “Claude Code waits in the open session and continues the task on its own after the limit resets.”

Price: Claude’s pricing says “Claude Code is included in all paid plans,” and the Free plan leaves it out. Pro is $20 a month, or $17 billed yearly; Max 5x is $100 and Max 20x is $200. Fable 5.1, which its docs count among “the most capable models in Claude Code,” comes with Max, which has “a separate weekly limit for Fable,” and on Pro it bills to usage credits.

Watch for: the limits are shared with your Claude chats: “all activity in both tools counts against the same limits.” Anthropic publishes no message counts. Its Pro plan page says “The number of messages you can send will vary,” and adds that “we may limit your usage in other ways, such as weekly and monthly caps.” And its top score is slow and dear: 61 minutes and $11.42 a task, against 24 minutes and $1.50 for Codex with GPT-6.1 Sol. Claude Code isn’t open source: its license reads “© Anthropic PBC. All rights reserved.” And on native Windows, its setup guide lists sandboxing as “Not supported”; WSL 2 supports it.

Codex

Best for: a lower cost per task, a free start, or an agent whose code you can read.

OpenAI’s Codex runs in the ChatGPT desktop app, in a terminal, in your editor and in OpenAI’s cloud. The Codex CLI is “a coding agent from OpenAI that runs locally on your computer,” open source under Apache 2.0, for macOS, Linux and Windows. The desktop app is for “macOS, Windows, or Linux,” in preview on Linux, and on Windows it “runs natively on Windows using PowerShell and the Windows sandbox.”

On the index, Codex with GPT-6 Astra scores 54.7 at $7.04 a task, and with GPT-6.1 Sol 54.1 at $1.50: 7.6 points behind Claude Code with Opus 5.5, for less than a seventh of the cost per task. GPT-6.1 Terra is cheaper still, at 48.2 for $0.69.

Price: “ChatGPT Work and Codex are included in your ChatGPT Free, Go, Plus, Pro, Business, Edu, or Enterprise plan,” says Codex’s pricing. Free, and Go at $8 a month, get “GPT-6 Luna at Standard speed in the desktop app, subject to rollout.” Plus, at $20, adds “Codex on the web, in the CLI, in the IDE extension, and on iOS” and GPT-6.1 Sol. OpenAI estimates that Plus allows 15 to 160 local messages every five hours with GPT-6.1 Sol, 35 to 285 with Terra and 5 to 45 with Astra. Pro is $100, $200 or $500 a month, and “Pro plans currently have no five-hour limit.”

Watch for: OpenAI says its counts “are not fixed message limits,” and “Weekly limits may also apply.” Codex shares them with ChatGPT Work, which “uses the same pricing, credits, and usage limits as Codex.” Codex with GPT-6 Luna, the free plans’ model, scores 31.1 on the index even at max effort, far below Sol. And on October 30, existing Pro 200 plans drop to a lower allowance: OpenAI says that after October 29, “your subscription will move to the lower included usage allowance.”

Both, each in its own app

Best for: running out of usage, or wanting a second lab’s view of the same code.

Anthropic’s and OpenAI’s limits are separate, so when one agent runs out, the other still works, and each can review what the other wrote. One instruction file covers both: Codex reads AGENTS.md, and Claude Code “can read AGENTS.md as your project instructions, so a repository already set up for other coding agents works without adding a CLAUDE.md,” per its docs. Several apps run both side by side, each on its own task; our Conductor alternatives guide compares them.

Price: two plans, from $40 a month for Claude Pro and ChatGPT Plus, or $37 with Pro billed yearly.

Watch for: each agent keeps its own conversations. Codex doesn’t see what Claude Code was doing, so when you move a task across, you tell the other agent where things stand.

Enjoy

Best for: using both in one conversation, and keeping going when one runs out.

Enjoy runs Claude Code and Codex, along with Grok Build, Antigravity, Muse Code and local models through OpenCode, on Mac, Windows and Linux, signed in with the plans you already have. What it adds for these two:

• One conversation, either agent. Switch agents at any point: “The new agent gets the conversation so far, including replies from other agents, before it works on your message.” Plan with Claude Code, build with Codex, then have Claude Code review it.
• Limits handled. When an agent runs out, Enjoy switches to another of your accounts with usage left, or waits for the reset and carries on. Select Stop waiting to hand the conversation to the other agent instead.
• One set of instructions. Project instructions apply “to every conversation in the project, whichever agent you use.”
• Usage in view. It shows what every signed-in Claude and ChatGPT account has left, and on Max it says when Max 5x would have been enough.

Price: Enjoy is free on your computer, with no account. Enjoy Cloud, a monthly subscription, adds your phone and any browser, your other computers and teammates. See pricing.

Watch for: Enjoy doesn’t add Opus to a ChatGPT plan or GPT to a Claude one: each agent runs on its own lab’s plan, and Enjoy runs the Codex CLI, which OpenAI includes from Plus. Waiting for a reset needs “Your computer must be on, with Enjoy open, when the usage resets,” and scheduled workflows also run only while it’s awake with Enjoy open. Enjoy isn’t open source, and the phone, other computers and teammates need Enjoy Cloud.

When one agent is enough

If you already pay for one and rarely run out, a second plan buys little. Claude Code waits out a session limit by itself, and both labs sell extra usage: “ChatGPT Plus and Pro users who reach their usage limit can purchase additional credits,” and Claude’s usage credits “are billed at standard API rates,” as our usage limit guide explains.

Move from Codex to Claude Code when the work is hard enough that the index’s 7.6-point gap shows up in your reviews. Move from Claude Code to Codex when its limits or cost are what stop you, or when you want an agent you can read and change.

How we compared

We read Anthropic’s pricing, help center, Claude Code docs and license, OpenAI’s Codex and ChatGPT pricing, help center, docs and GitHub repository, and Artificial Analysis’s index on October 27, 2026, and every one is linked below. We quote each lab’s own words for its limits and leave out what it hasn’t documented. A benchmark scores tasks, not how far a plan goes, so we give the limits each lab publishes beside it. Plans and models change often, so we recheck this guide against its sources at least once a month, and the date at the foot of the page says when we last did.