Two launches, one tie
On 22 September Anthropic released Claude Opus 5.5 with a table showing it at 66.4% on Terminal-Bench 4.0, a test of agents doing real work in a terminal, against 57.9% for OpenAI's GPT-6 Astra. The same day Artificial Analysis, which runs the benchmark itself, published its own figure for Opus 5.5: 59.6%, level with Astra.
Nobody's number was false. Anthropic ran the test in its own setup, and its footnotes say it used the xhigh effort setting for Opus 5.5 and quoted OpenAI's own figure for Astra at high, although OpenAI's model page lists two settings above high. Artificial Analysis ran Astra at xhigh in its own test setup; its write-up does not say which setting produced the Opus figure. Change who runs the test and how, and a clear win becomes a tie.
| Row | Launch table | Artificial Analysis run |
|---|---|---|
| Claude Opus 5.5 | 66.4% | 59.6% |
| GPT-6 Astra | 57.9% | 59.6% |
That is the right way into Claude Code against Codex this autumn. The two products have grown into near copies of each other: the same $20, $100 and $200 price points (OpenAI has since added a $500 plan), the same spread across terminal, editor, desktop app and cloud, the same parallel sessions and support for the Model Context Protocol. What separates them now is smaller and more practical: which model each hands you by default, how much it tells you about your limits, how far its plans stretch, and how it treats the machine it runs on. On those terms Claude Code gives you the stronger default model. Codex gives you a wider range of plans, limits you can read in advance, an open-source CLI and better footing on Windows.
Four releases since that launch day have moved the details. Anthropic shipped Claude Sonnet 5.5 on 28 September and Claude Haiku 5.5 on 7 October. OpenAI released GPT-6.1 Sol at DevDay on 29 September, which the Codex CLI made its default the same day, and added GPT-6.1 Terra to Codex on 26 October. In September Codex also owned the cheap end of the model range; Haiku 5.5, priced the same as GPT-6 Luna, ended that, and it is the one change to our verdict. Anthropic has also renamed its $200 plan, as the plan table below shows. Our comparison of Claude Code, Codex, Cursor, Antigravity and GitHub Copilot sets these two against the editors, and our Cursor vs Antigravity comparison covers the editors in detail.
The products have converged
The old shorthand, Claude Code in your terminal and Codex in OpenAI's cloud, no longer holds. Anthropic's Claude Code documentation lists a terminal CLI, extensions for VS Code and JetBrains, a desktop app and a web version whose cloud sessions keep running after you close the laptop. OpenAI's Codex documentation lists a CLI, an IDE extension, the desktop app and Codex cloud, which runs tasks in parallel and hands back each one as it reaches a reviewable result. DevDay, according to OpenAI's recap, added voice control in the CLI, code review in the desktop app with automatic first-pass reviews in the cloud, and Codex Security Cloud, a research preview that scans connected GitHub repositories and watches new commits.
The plans converged too. Both companies fold the coding agent into the general subscription rather than selling it separately, and the tiers line up almost exactly:
| Monthly price | Anthropic (Claude Code) | OpenAI (Codex) |
|---|---|---|
| $0 | Free: no Claude Code | Free: GPT-6 Luna in the desktop app, subject to rollout |
| $8 | No equivalent | Go: same Codex access as Free |
| $20 | Pro | Plus |
| $100 | Max: five times Pro, per five hours and per week | Pro $100, no five-hour limit |
| $200 | Ultra (formerly Max 20x): ten times Pro per week, no five-hour limit once verified | Pro $200, open to new subscribers again with a lower allowance |
| $500 | No equivalent | Pro $500, the only plan with Ultrafast |
| Per user, teams | Team standard seat, $20 billed annually or $25 monthly | Business, $20 billed annually or $25 monthly |
Sources: Claude pricing, Anthropic's help pages on the Max and Ultra plans, OpenAI's Codex pricing page and its Pro tiers page, as listed on 3 November 2026.
The asymmetry is at the top, and it has flipped since September. On 10 September OpenAI stopped taking new sign-ups and upgrades for its $200 tier, citing demand for Astra, TechCrunch reported. That pause is over. OpenAI's Pro tiers page says the $200 plan is open to new subscribers again with a lower allowance (earlier subscribers kept the old one until 29 October), and a new $500 plan is the only one with Ultrafast, which OpenAI says generates tokens up to eight times faster in Codex. Anthropic's ladder still stops at $200, where it renamed Max 20x as Ultra. The help page describes Ultra as ten times Pro's weekly usage, with the five-hour limit removed for accounts Anthropic has verified from their account and payment history.
The default model is the biggest difference
Set the two model lineups side by side and, since Haiku 5.5, they match at both ends and differ only in the middle.
| Tier | Anthropic, per million tokens in / out | OpenAI, per million tokens in / out |
|---|---|---|
| Top | Claude Fable 5.1: $10 / $50 | GPT-6 Astra: $10 / $50 |
| Workhorse | Claude Opus 5.5: $4 / $20 | GPT-6.1 Sol: $2 / $10 |
| Mid | Claude Sonnet 5.5: $2 / $10 | GPT-6.1 Terra: $1 / $5 (since 26 October) |
| Small | Claude Haiku 5.5: $0.10 / $0.50 | GPT-6 Luna: $0.10 / $0.50 |
| Context window | 1M tokens on all four | 1.05M tokens (GPT-6.1 Sol) |
Figures from Anthropic's models overview and OpenAI's API pricing and model pages. OpenAI charges more for prompts beyond 272K input tokens, and Haiku 5.5's price applies to prompts up to 100,000 tokens.
What matters day to day is which row each tool starts you on. Claude Code's model configuration page says the default now resolves to Opus 5.5 on Pro, Max, Ultra, Team, Enterprise and the API; before version 2.1.280, Pro users defaulted to Sonnet 5. Opus 5.5 runs at medium effort unless you change it, and on Anthropic's API the sonnet and haiku aliases now resolve to Sonnet 5.5 and Haiku 5.5. The Codex CLI has defaulted to GPT-6.1 Sol since 29 September, and OpenAI's models page describes it as "near-Astra performance for complex work at a lower cost than Astra." GPT-5.5 was retired from Codex for ChatGPT sign-ins on 14 October. Our piece on Claude Opus 5.5 for SEO and Google Ads covers what the Anthropic model changed beyond coding.
So an untouched Claude Code session still runs a model that costs twice as much per token as the one an untouched Codex session runs, and the independent numbers still favor it, by less than in September. Version 5 of the Artificial Analysis Intelligence Index, introduced on 27 October, scores Opus 5.5 at 58 and GPT-6.1 Sol at 55, both at maximum effort. At medium, Claude Code's default, Opus 5.5 scores 48, and GPT-6.1 Sol at medium scores 46.
| Row | Medium effort | Maximum effort |
|---|---|---|
| Claude Opus 5.5 (Claude Code default) | 48 | 58 |
| GPT-6.1 Sol (Codex default) | 46 | 55 |
In September a Codex user who wanted Opus-class results had to pick Astra, which costs two and a half times as much as Opus 5.5 per token and, on Plus, comes with 5 to 45 messages per five-hour window. OpenAI now says GPT-6.1 Sol matches Astra on DeepSWE v1.1, a test in real codebases, at about a fifth of the cost. That is a vendor result on vendor settings, but it weakens the case for paying Astra's price.
The middle row has new models on both sides. Sonnet 5.5 replaced Sonnet 5 at the same price, which is also GPT-6.1 Sol's price, and the new index puts it at 54. OpenAI's GPT-6.1 Terra costs half that, scores 51 at maximum effort, and is pitched by OpenAI at drafting and everyday work rather than hard coding; on Plus it allows 35 to 285 messages per five hours. Anthropic reports Sonnet 5.5 at 70.6% on Terminal-Bench 4.0, ahead of Opus 5.5's 66.4%, and Artificial Analysis also found it ahead, 64% to 60%, but measured it using about 193,000 output tokens per task on its index, the most it has recorded. A cheaper token does not guarantee a cheaper job.
Both companies fence off their most expensive model. Anthropic's help center says Fable 5.1 is outside Pro's usage limits and available there only through paid usage credits; on Max and Ultra, you can spend up to half your weekly allowance on Fable models. OpenAI includes Astra on Plus but gives it the smallest allowance of any model.
At the bottom the gap has closed. In September GPT-6 Luna cost a tenth of Claude Haiku 4.5 per token; Haiku 5.5, released on 7 October, costs the same $0.10 and $0.50 as Luna. Neither is a strong coder on its own: the new Artificial Analysis index scores Luna 28 and Haiku 5.5 36 at maximum effort, against 55 for GPT-6.1 Sol. These are the models for renames, lint fixes and other edits that need speed more than judgment. What Codex keeps at this end is a published allowance: a Plus subscriber gets 350 to 3,000 Luna messages per five hours, and Anthropic publishes no count for Haiku 5.5.
What the limits look like
OpenAI publishes its Codex limits as numbers. The pricing page gives ranges per model for each five-hour window: on Plus, 15 to 160 messages with GPT-6.1 Sol, 5 to 45 with Astra and 350 to 3,000 with Luna. Pro, at $100, $200 or $500, currently has no five-hour limit, though other limits may apply. The ranges are wide because a message that reads half a repository and reasons for minutes costs far more than one that fixes a typo, but they give a buyer something to plan against.
Anthropic publishes multiples instead. Max gives five times Pro's usage every five hours and every week; Ultra gives ten times Pro's weekly usage and, once the account is verified, no five-hour limit at all. Claude Code draws on the same limits as everything else you do in Claude. There is no published message count for Pro, so a week of real use is the only way to find out whether it is enough.
Both companies sweetened their September launch days the same way. Anthropic raised five-hour limits on Pro, Max, Team and seat-based Enterprise plans and added a rate-limit reset you can save and spend when you choose. OpenAI's announcement gave Plus, Pro and Business users a "banked reset".
Cloud work is not extra capacity on either side. Claude Code's cloud documentation says cloud sessions share rate limits with the rest of your account, parallel sessions use them up proportionately, and there is no separate charge for the VM. OpenAI's pricing page says local messages and cloud chats share the plan's allowance. If you prefer to pay per token, both run on API billing: Claude Code with an Anthropic Console account, Codex with an API key at standard rates.
How each one treats your machine
An agent that asks permission for every shell command is tiring to use; one that asks for none is a risk. Both tools resolve this the same way, with an operating-system sandbox that fences off files and network so the agent can run routine commands unattended. The implementations differ where it counts for some teams.
Claude Code's sandbox uses macOS's built-in Seatbelt framework and, on Linux and WSL2, two extra packages. Its documentation says native Windows is not supported: Claude Code itself runs on Windows, but to get the sandbox you run it inside WSL2. Codex's sandbox uses Seatbelt on macOS, bubblewrap on Linux and WSL2, and a native Windows sandbox in PowerShell. Its defaults let the agent edit files inside the project and run routine commands, and make it stop and ask before going further, including onto the network. For a Windows shop that does not use WSL, that is a practical reason to prefer Codex.
In the cloud, each Claude Code session runs in an isolated virtual machine that Anthropic manages, with network access limited by default, git credentials kept outside the sandbox, and GitHub required for cloning and opening pull requests. Codex cloud runs tasks in isolated environments you configure per repository, and you can start them from GitHub pull requests, GitLab merge requests and issues, Linear issues or Slack threads. Claude Code covers similar ground through Slack, GitHub Actions and GitLab CI/CD, cloud scheduled tasks (which Anthropic used to call routines) that fire on a schedule or on GitHub events, and an auto-fix mode that watches a pull request and pushes fixes for failed checks and review comments.
For parallel work on one machine, Claude Code's background agents move each session into its own git worktree before editing, so several can read one checkout without overwriting each other. Codex supports worktrees too, and its Ultra setting splits a single task among subagents.
Claude Code has the deeper kit for packaging repeatable work: CLAUDE.md for standing project instructions, skills for workflows a team shares (our guide to Claude skills for SEO automation shows several), and hooks that run shell commands before or after the agent acts. Codex reads AGENTS.md and supports MCP servers, as Claude Code does. Claude Code will also read an existing AGENTS.md, so one instruction file can serve both agents in the same repository.
One difference is structural. The Codex CLI is open source, developed in public on GitHub. A team that wants to read, audit or patch the agent running on its developers' machines can do that with Codex.
Reading the launch tables
Go back to the table that opened this piece. OpenAI's own Astra announcement on 3 September reported 57.9% on Terminal-Bench 4.0 against 55.8% for Claude Fable 5.1. Three weeks later Anthropic's table had Opus 5.5 at 66.4%, with Astra run at high. Artificial Analysis found them level at 59.6%. Opus 5.5 scored almost seven points lower in the independent run than in Anthropic's, so the test setup matters as much as the model. Each vendor's figure is a real result under conditions the vendor chose; the independent run is the only one of the three that held the conditions constant for both models.
Two more reasons to hold launch tables loosely. First, benchmark runs use high or maximum effort, and your tool probably does not: Claude Code runs Opus 5.5 at medium by default, and OpenAI tells Codex users to start GPT-6.1 Sol at whatever effort their client sets by default. Second, cost per task depends on how many steps an agent takes as well as its price per token. Artificial Analysis's leaderboard lists what each model cost per task to run its index: $2.45 for Opus 5.5 at medium and $10.04 at maximum, against $0.36 and $1.27 for GPT-6.1 Sol. Opus costs twice as much per token and, on those figures, about seven to eight times as much per task. Those are model runs through the API, not sessions inside Claude Code or Codex, but they show how far token counts per task can drift from the price list.
Anthropic's most eye-catching claim, that one tester finished a 680,000-line code migration in under a day, is a single unnamed example. It may be true. It is not something to plan a quarter around.
The comparison that settles it for a team is its own. Take five tasks from your backlog, run each through both tools at their default settings for a week, and count what merged, how long review took, and whether anyone hit a limit.
Which fits which work
For a long, tightly coupled change in one codebase, where step nine depends on what happened at step three (a framework migration, a refactor across dozens of files), start with Claude Code. Its default model is the stronger of the two defaults, and background agents in separate worktrees let you split off side tasks without disturbing the main job.
For a steady flow of small, well-specified tickets, or mechanical changes across many files, Codex is still the easier tool to plan around, though no longer clearly the cheaper one: Haiku 5.5 now costs what Luna costs per token, so Codex's edge here is the published limits that tell you in advance how far a plan will stretch. Codex also remains the pick if your developers work on native Windows or your security team wants to read the agent's source.
If your team repeats the same jobs every week (a release checklist, a dependency audit, a report built from the same exports), Claude Code's skills, hooks and cloud scheduled tasks make those repeatable in a way Codex does not yet match.
Running both is common and cheap to set up. Put shared instructions in AGENTS.md, give each agent its own branch or worktree, and review everything as a pull request.
Search and ads teams face the same split. A coding agent is the right tool for site-wide template, markup and redirect changes; the numbers that justify those changes should come from code reading Search Console and Google Ads exports, with a person approving the result. Our guide to using Claude Code for SEO works through seven such jobs. That division is how AIO Copilot runs SEO and Ads work for businesses and agencies. It is priced per site; tell us how many you manage and we email the price.
Both tools' $20 tiers include a model near the top of the independent tables. A week on each at the lowest paid tier will tell you more than any launch table, and with one AGENTS.md file serving both, the second trial costs little beyond the subscription. Buy a $100 tier or higher only after you have hit the $20 limits doing real work.
Questions
Is Claude Code or Codex better in November 2026?
Neither wins everywhere. Claude Code starts you on Claude Opus 5.5, which version 5 of the Artificial Analysis Intelligence Index scores at 58 against 55 for GPT-6.1 Sol, the Codex default since 29 September, both at maximum effort. Codex offers a wider range of plans, from free to $500 a month, published usage limits, an open-source CLI and a native Windows sandbox. Pick by the work you do most, then test both on your own repository.
How much does Claude Code cost?
Claude Code is included in Claude Pro at $20 a month, Max at $100 and Ultra (formerly Max 20x) at $200, and in Team seats at $25 or $125 a month ($20 or $100 billed annually) and Enterprise seats; the Free plan does not include it. Through the API, the default model, Claude Opus 5.5, costs $4 per million input tokens and $20 per million output tokens, Sonnet 5.5 costs $2 and $10, and Haiku 5.5 $0.10 and $0.50 for prompts up to 100,000 tokens.
How much does Codex cost?
Codex comes with ChatGPT plans rather than as a separate subscription: Free and Go ($8) with GPT-6 Luna, Plus at $20 a month, Pro at $100, $200 or $500, and Business at $20 per user a month billed annually or $25 monthly. Only the $500 Pro plan gets the Ultrafast speed tier. Through the API, GPT-6.1 Sol costs $2 and $10 per million input and output tokens, GPT-6.1 Terra $1 and $5, GPT-6 Luna $0.10 and $0.50, and GPT-6 Astra $10 and $50.
What is the default model in Claude Code and in Codex?
Claude Code defaults to Claude Opus 5.5 on Pro, Max, Ultra, Team, Enterprise and API accounts, at medium effort, and its sonnet and haiku aliases now point to Sonnet 5.5 and Haiku 5.5. The Codex CLI has defaulted to GPT-6.1 Sol since 29 September 2026, the day OpenAI released it at DevDay, and GPT-5.5 was retired from Codex for ChatGPT sign-ins on 14 October. Both let you switch models and effort per session.
Can I use Claude Code and Codex in the same repository?
Yes. Codex reads an AGENTS.md instruction file and Claude Code can read AGENTS.md as well as its own CLAUDE.md, so one set of project instructions can serve both. Give each agent its own branch or git worktree so their edits do not collide.
New studies, by email
Our own studies, and what each new release changes for SEO, AEO and Google Ads.
