For a UK team of 5 to 50 developers, Claude Code and OpenAI Codex are close enough on features that the decision usually comes down to admin controls, data residency and which models your team prefers. On list prices checked on 5 November 2026, a heavy-use seat costs £90 a month on Claude Team Premium and £75 a month on ChatGPT Business Premium, and neither business plan trains on your code by default.
A disclosure first. Kelriva builds with Claude Code every day, and I hold Anthropic's Claude Code 101 certificate. That gives us a bias, so every claim below comes from the vendors' own pricing pages and documentation, checked today. Where Codex is the better fit, I say so.
What each tool is
Claude Code is Anthropic's coding agent. It reads your codebase, edits files, runs commands and opens pull requests. Anthropic's docs list a terminal CLI, extensions for VS Code, Cursor and JetBrains IDEs, a desktop app, a web version at claude.ai/code and the Claude mobile apps. It also runs in GitHub Actions and GitLab CI/CD, offers automatic GitHub code review, and takes tasks from Slack. It is included in every paid Claude plan except Free.
Codex is OpenAI's coding agent, and it now sits inside ChatGPT. OpenAI's docs list the Codex CLI, an IDE extension, Codex cloud on the web, the ChatGPT desktop and mobile apps, and an SDK for scripted work. Cloud features include delegating GitHub issues and pull requests by tagging @codex, automatic PR reviews, and Slack and Linear integrations. The Codex CLI is open source, at github.com/openai/codex. Codex usage is shared with ChatGPT Work, OpenAI's agent for non-coding tasks.
Both tools are agents, not autocomplete. You give them a task, they plan, change several files, run tests and come back with a diff. Both support MCP for connecting internal tools, skills, subagents and project instruction files.
Side-by-side comparison
| Claude Code | OpenAI Codex | |
|---|---|---|
| Terminal | Yes, Claude Code CLI | Yes, Codex CLI (open source) |
| IDE | VS Code, Cursor, JetBrains | IDE extension |
| Web and cloud agent | Claude Code on the web, cloud sessions in Anthropic-managed VMs | Codex cloud, cloud chats and environments |
| Desktop and mobile | Desktop app, iOS and Android | ChatGPT desktop app, iOS |
| GitHub | GitHub Actions, GitHub Code Review, GitLab CI/CD | @codex delegation on issues and PRs, automatic PR reviews |
| Chat tools | Slack | Slack, Linear |
| Current models | Fable 5.1, Opus 5.5 (default on Team, Enterprise and API), Sonnet 5.5, Haiku 5.5 | GPT-6 Astra, GPT-6.1 Sol, GPT-6.1 Terra, GPT-6 Luna |
| Team plan, per seat per month | Team Standard £18, Premium £90 (billed annually) | ChatGPT Business Standard £15, Premium £75 (billed annually) |
| Enterprise | US$20 a seat a month plus API-rate usage, 20-seat minimum | Custom pricing |
| Usage limits | Per-seat allowance on a rolling five-hour window and a weekly window, shared with Claude chat; Premium is 5x Standard | Allowance per five-hour window, weekly limits may apply, shared with ChatGPT Work; Business Premium has no five-hour limit |
| Going past the limit | Usage credits with spend limits per organisation, group or member | ChatGPT credits on flexible pricing |
| Team admin | SSO, usage analytics, spend limits, managed settings | SAML SSO, MFA, managed config policies, usage analytics and spend controls |
| Enterprise admin | Adds SCIM, audit logs, Compliance API, custom retention, role-based access | Adds SCIM, EKM, RBAC, Compliance API and audit logs, retention and residency controls |
| Training on business data | None by default on Team, Enterprise and API | None by default on Business, Enterprise and API |
| Pay per token instead | Claude API key or Bedrock, Google Cloud, Microsoft Foundry | OpenAI API key (CLI, SDK, IDE only, no cloud features) |
Claude prices are from our Claude Code cost guide, which lists the UK figures on claude.com/pricing. Anthropic's UK page says its plan prices include 20% VAT. ChatGPT Business prices are from OpenAI's UK business pricing page today. They are the annual-billing rates: the same page lists US$20 and US$100 a seat billed annually, or US$25 and US$125 billed monthly, and does not show a monthly price in pounds.
Models and API pricing
On Claude Code, Opus 5.5 is the default model on Team, Enterprise and API accounts, according to Anthropic's model configuration docs. You can switch to Sonnet 5.5 for everyday work, Haiku 5.5 for quick subagent tasks, or Fable 5.1 for the hardest problems.
On Codex, OpenAI describes GPT-6 Astra as its most capable model, GPT-6.1 Sol as near-Astra performance at lower cost, GPT-6.1 Terra as fast and efficient, and GPT-6 Luna as the most efficient for focused, high-volume tasks. GPT-5.5 was retired from Codex for ChatGPT sign-ins on 14 October 2026.
| Model | Input | Cached input | Output |
|---|---|---|---|
| Claude Fable 5.1 | $10 | $0.25 | $50 |
| Claude Opus 5.5 | $4 | $0.20 | $20 |
| Claude Sonnet 5.5 | $2 | $0.10 | $10 |
| Claude Haiku 5.5 (prompts up to 100k tokens) | $0.10 | $0.01 | $0.50 |
| GPT-6 Astra | $10 | $1.00 | $50 |
| GPT-6.1 Sol | $2 | $0.10 | $10 |
| GPT-6.1 Terra | $1 | $0.05 | $5 |
| GPT-6 Luna | $0.10 | $0.01 | $0.50 |
The tiers line up closely. Sonnet 5.5 and GPT-6.1 Sol have the same list price, as do Haiku 5.5 and GPT-6 Luna. Price per token will not decide this for you. How many tokens each agent uses to finish your kind of task will, and only a pilot on your own code tells you that.
What a team of 10 developers costs
Most teams of this size buy the business plan, not the API. Developers who use an agent most of the day will hit a standard seat's limits, so price them on heavy-use seats.
| Option | Per seat | 10 seats a month | 10 seats a year |
|---|---|---|---|
| Claude Team Premium (billed annually) | £90 | £900 | £10,800 |
| ChatGPT Business Premium | £75 | £750 | £9,000 |
| Claude Team Standard (billed annually) | £18 | £180 | £2,160 |
| ChatGPT Business Standard | £15 | £150 | £1,800 |
| Claude via API, Anthropic's enterprise average | $150 to $250 | $1,500 to $2,500 | $18,000 to $30,000 |
The API row uses Anthropic's own figure: across enterprise deployments, Claude Code averages around $13 per developer per active day and $150 to $250 per developer per month. OpenAI does not publish an equivalent per-developer figure for Codex, and its docs warn against using API token prices to estimate subscription usage, so I have left that row out rather than guess.
A mix is often cheaper. Give heavy users premium seats and everyone else standard seats; both vendors let you mix seat types. Both also sell credits for usage beyond the allowance, so set spend limits on day one. Note that OpenAI says that from December, plan usage will share one limit across all of ChatGPT and Codex, so a developer's chat use will eat into their coding allowance. Claude Team already works this way: Claude Code shares one pool with Claude chat.
Data and UK GDPR
On training, the two vendors now say the same thing for business accounts. Anthropic's Claude Code docs say it does not train generative models on code or prompts sent under commercial terms, which covers Team, Enterprise and the API. OpenAI says it does not use data from ChatGPT Business, Enterprise or its API platform for training by default. Personal plans are different on both sides: Claude Pro, Max and Ultra, and ChatGPT Plus and Pro, use an opt-out setting. Company code belongs on company accounts.
Retention and residency are where they differ. Anthropic's standard retention for commercial Claude Code users is 30 days, with zero data retention available to qualified Enterprise accounts. Claude Code also keeps session transcripts on each developer's machine, in plain text, for 30 days by default, which your laptop security policy should cover. OpenAI lists data residency in ten regions, including the UK and Europe, for eligible Enterprise, Edu and API customers, but not for ChatGPT Business. I did not find a UK residency option on the Anthropic pages I checked; Anthropic documents a US-only inference option at a 1.1x price.
For UK GDPR, both are processors on business terms, and you still need a data processing agreement, a record of where code and prompts go, and a rule about secrets and personal data in repositories. Our guide to Claude, UK GDPR and client data covers the options in more detail, and most of it applies to Codex too.
Which to choose
Choose Codex if your company already runs on ChatGPT Business or Enterprise. Your developers get Codex inside seats you are paying for, and admins manage one workspace. Codex is also the stronger pick if you need UK data residency and qualify for ChatGPT Enterprise, or if your workflow lives in Linear, which Codex integrates with directly.
Choose Claude Code if your team wants Anthropic's models, or if you want to run the agent through your own cloud account. Claude Code supports Amazon Bedrock, Google Cloud and Microsoft Foundry, so billing and data stay inside a cloud contract you already have. Anthropic also documents a JetBrains plugin alongside VS Code and Cursor. If half your team lives in IntelliJ or PyCharm, check each vendor's IDE docs before you choose.
Choose on admin needs if you are near 20 seats. Claude Enterprise starts at 20 seats and adds SCIM, audit logs and the Compliance API. ChatGPT Business tops out at 200 people, and Enterprise adds SCIM, RBAC and residency. If your security team needs audit logs, budget for an enterprise tier on either side.
If you can't decide, run both. Give five developers a month on each, on the same backlog, and compare merged pull requests, review time and spend. Codex reads AGENTS.md for project instructions, and Claude Code can read AGENTS.md in place of CLAUDE.md, so one instruction file serves both tools. Plenty of teams keep both for good, using one for cloud tasks and the other in the terminal. If you go that way, our note on model-agnostic infrastructure explains how to avoid getting locked in.
How to run a fair four-week pilot
Pick tickets from your real backlog, not demo tasks. Split them evenly by size and type, so each tool gets bug fixes, small features and at least one refactor. Use business seats from day one, so the data terms match what you would roll out. Set a spend limit on each side and check the usage dashboards weekly. Both vendors say usage varies a lot with model choice and context, so the first week is often unrepresentative.
At the end, compare four things: pull requests merged without major rework, reviewer time per pull request, total spend including credits, and what developers say they would keep. The last one matters more than it sounds. A tool your team avoids is the expensive one, whatever its seat price.
How Kelriva can help
If you want a second opinion on rolling out a coding agent, including seat mix, spend limits and the data processing paperwork, we can help as part of an AI Readiness Assessment, from £3,500 over 9 to 14 days. Or start with our free 2-minute AI readiness score.
Sources
All links checked on 5 November 2026. Claude plans and pricing. Claude Code overview. Claude Code: manage costs. Claude Code: data usage. Claude Code: model configuration. Claude API pricing.
ChatGPT pricing. ChatGPT Business pricing. Codex pricing and plan features. Codex models. OpenAI API pricing. OpenAI business data privacy. Codex open source.