Last updated: November 7, 2026
TL;DR
Claude Code and Codex cost the same to start: $20 a month for Claude Pro or ChatGPT Plus, and both companies sell plans at $100 and $200. The real difference in Claude Code vs Codex pricing is what each company tells you you're buying. Anthropic states every plan's usage as a multiple of Pro, so you can work out the cost per unit of usage. OpenAI gives message estimates for Plus and lists three Pro tiers, but its Codex pricing page doesn't say how much more usage each Pro tier includes.
- 💵 At $20. Claude Pro ($17 a month billed annually) and ChatGPT Plus both include their coding agent in the terminal, the IDE and the web.
- 📈 Above $20. Claude Max is $100 for 5x Pro's usage and Ultra is $200 for 10x Pro's weekly usage. ChatGPT Pro comes at $100, $200 or $500 with no five-hour limit.
- 🧮 Per unit of usage. Every Claude personal plan works out to about $20 per Pro's worth of weekly usage, or $17 on annual Pro. Bigger plans buy room, not a discount.
- 🔑 Pay as you go. Claude Code defaults to Opus 5.5 at $4 per million input tokens and $20 per million output. Codex's GPT-6.1 Sol costs $2 and $10 through the API.
- 🧭 How to choose. Start on the $20 plan of the tool you prefer, watch where you hit limits for two weeks, and move up only when the limit costs you work.
Claude Code vs Codex pricing, plan by plan
Both tools come bundled with a chat subscription, so the price you compare is the plan, not the coding agent. Here is what each company lists, from Claude's pricing page and OpenAI's Codex pricing page, as of November 6, 2026:
| Price per month | Claude plan | What Anthropic says about usage | ChatGPT plan | What OpenAI says about Codex usage |
|---|---|---|---|---|
| $0 | Free | No Claude Code | Free | GPT-6 Luna in the desktop app, subject to rollout |
| $8 | – | – | Go | GPT-6 Luna in the desktop app, subject to rollout |
| $20 ($17 annual on Claude) | Pro | The baseline: 1x every 5 hours and every week | Plus | About 15 to 160 local messages per five hours on GPT-6.1 Sol, weekly limits may apply |
| $100 | Max | 5x Pro every 5 hours and every week | Pro | No five-hour limit |
| $200 | Ultra | 10x Pro every week, no 5-hour limit for verified accounts | Pro | No five-hour limit |
| $500 | – | – | Pro | No five-hour limit, plus Ultrafast mode |
Claude Code is included in every paid Claude plan and shares one usage pool with chat: "your work in the terminal and your chats draw from one pool," per Claude's pricing FAQ. Codex works the same way on the ChatGPT side, where ChatGPT Work and Codex share usage, and local messages and cloud chats draw from one allowance (OpenAI).
Max and Ultra are billed monthly only. Pro is the only individual Claude plan with an annual price.
At $20: Claude Pro vs ChatGPT Plus
At the entry price the two plans are close, and the choice comes down to the tool you'd rather use. Claude Pro costs $17 a month billed annually ($200 up front) or $20 monthly. ChatGPT Plus costs $20 a month.
| Claude Pro | ChatGPT Plus | |
|---|---|---|
| Price | $17 annual, $20 monthly | $20 monthly |
| Coding agent | Claude Code, included | Codex on the web, in the CLI, the IDE extension and on iOS |
| Default model | Opus 5.5 (Claude Code docs) | GPT-6.1 Sol and GPT-6 Luna are included |
| Usage stated as | The baseline every other Claude plan is measured against | Local messages per five hours, per model |
| When you run out | Usage credits at standard API rates | Buy credits, or use an API key at API rates |
OpenAI's per-model estimates for Plus are ranges, not fixed limits: 15 to 160 local messages per five hours on GPT-6.1 Sol, 5 to 45 on GPT-6 Astra and 350 to 3,000 on GPT-6 Luna. Anthropic gives no message count at all. Its pricing FAQ says usage "depends on the length and complexity of your conversations, the model you choose, and the features you use, so there's no fixed message count."
So neither $20 plan tells you how many coding sessions you'll get. Both draw down faster with bigger models, longer context and more tool calls.
At $100 and $200: Claude Max and Ultra vs ChatGPT Pro
Above $20, Anthropic tells you the size of what you're buying and OpenAI tells you the price. Claude Max is $100 a month for 5x Pro's usage every 5 hours and every week. Ultra is $200 for 10x Pro's weekly usage, with no 5-hour limit for verified accounts. Both include Fable, up to half your weekly limit, and monthly API credits to build on the Claude Platform: $100 on Max and $200 on Ultra.
ChatGPT Pro comes in three tiers, at $100, $200 and $500 a month. OpenAI's Codex pricing page says "Pro plans currently have no five-hour limit," that Ultrafast mode comes with Pro $500, and that every Pro tier includes everything in Plus. ChatGPT's own pricing page adds monthly API credits to each Pro tier and "Longer Codex and ChatGPT Work sessions." How much more usage each tier holds than Plus isn't on the Codex pricing page. It links to a separate help article for that comparison.
That gap matters for the math. With Anthropic's multiples, you can price a unit of usage:
| Claude plan | Price per month | Weekly usage vs Pro | Cost per Pro's worth of weekly usage |
|---|---|---|---|
| Pro, annual | $17 ($200 up front) | 1x | $17 |
| Pro, monthly | $20 | 1x | $20 |
| Max | $100 | 5x | $20 |
| Ultra | $200 | 10x | $20 |
Every monthly Claude plan lands on $20 per Pro's worth of weekly usage. A bigger plan isn't cheaper usage. It buys room: a larger 5-hour window on Max, no 5-hour window on Ultra, Fable included, and API credits. The Claude Team vs Pro guide runs the same math for Team seats.
The same table can't be built for ChatGPT Pro from OpenAI's Codex pricing page, because the page doesn't state each tier's usage as a multiple of Plus. If you're comparing $100 plans, open the help article OpenAI links from its pricing page, then watch your own usage dashboard. In the Codex CLI, /status shows your remaining limits during a session.
Pay as you go: API rates for each default model
Both tools also run on an API key instead of a subscription, billed per token. Claude Code bills by API token use through a Console account, and Codex with an API key runs in the CLI, the SDK and the IDE extension, without cloud features such as GitHub code review or Slack.
| Claude Code default (Opus 5.5) | Claude Sonnet 5.5 | Codex GPT-6.1 Sol | |
|---|---|---|---|
| Input, per million tokens | $4 | $2 | $2 |
| Output, per million tokens | $20 | $10 | $10 |
| Cached input, per million tokens | $0.20 | $0.10 | $0.10 |
| Long prompts | Same rate across the full 1M context | Same rate across the full 1M context | $4 input and $15 output above 272K input tokens |
Prices are from Anthropic's API pricing page and OpenAI's API pricing page. Per token, Codex's GPT-6.1 Sol costs half of Opus 5.5 and the same as Sonnet 5.5, up to 272K input tokens.
Per-token price isn't per-task price, though. Two agents can use very different numbers of tokens for the same job, and OpenAI warns that API token prices are separate from subscription usage, so they shouldn't be used to estimate included tasks.
The only spend figure either company publishes for real use is Anthropic's. Across enterprise deployments, Claude Code averages about $13 per developer per active day and $150 to $250 per developer per month, with 90% of users under $30 per active day (Claude Code costs). On those numbers, the average developer's API spend is above Max's $100 price, though Anthropic doesn't say how much API spend fits inside Max's limits.
When you hit the limit
Both companies let you keep working past the plan, at a price. On a paid Claude plan you can wait for the reset, move up a plan, or turn on usage credits "to keep working at standard API rates," per Claude's pricing FAQ.
On ChatGPT, Plus and Pro users can buy credits without upgrading, and anyone can run extra local chats with an API key at API rates (OpenAI). Codex credits are charged per million tokens. GPT-6.1 Sol costs 50 credits per million input tokens, 2.5 per million cached and 250 per million output. The page says credit purchase prices depend on your plan, so it doesn't convert credits to dollars.
Two things spend either allowance faster than the model choice alone:
- ⚡ Speed modes. Codex's Fast mode uses included subscription usage at 2.5x the Standard rate, and Ultrafast at 8x.
- 🧵 Parallel agents. In Claude Code, background sessions and subagents count toward the same limits as your main session, so ten background sessions use quota about ten times as fast as one (agent view docs). Our guide to running Claude Code agents in parallel covers when that's worth it.
For a team: Claude Team vs ChatGPT Business
For two or more people, the sticker prices match. A Claude Team Standard seat and ChatGPT Business both cost $20 per user per month billed annually and $25 billed monthly.
| Claude Team | ChatGPT Business | |
|---|---|---|
| Entry seat | Standard: $20 annual, $25 monthly, more usage than Pro | $20 annual, $25 monthly, for 2 or more users |
| Heavier seat | Premium: $100 annual, $125 monthly, 5x a Standard seat | – |
| Usage stated as | Standard is "more usage than Pro," Premium is 5x Standard | Standard Business has the same local-message estimates as Plus |
| Shared extras | Up to $500 a month in pooled API credits | Larger virtual machines for cloud chats |
Claude Team adds a Premium seat for anyone who runs Claude Code most of the day. ChatGPT Business with flexible pricing lets the workspace buy more credits instead. Our guide to Claude Team vs Pro vs Max walks through choosing a seat person by person.
How to choose
Choose the tool first and the plan second, because the entry prices are the same. Then size the plan from your own usage, not from a comparison table.
- Start on the $20 plan of the tool you'd rather work in. Claude Pro annual saves $3 a month if you're sure.
- Watch your limits for two weeks. Claude Code's
/usageshows what counts against your plan, and Codex's/statusshows your remaining limits. - If you hit the 5-hour limit most days, move up. Claude Max raises both windows 5x. Claude Ultra (for verified accounts) and every ChatGPT Pro tier drop the five-hour limit.
- If you hit limits only some weeks, keep the $20 plan and pay for the overflow with usage credits on Claude or credits on ChatGPT.
- For CI and scripts, use an API key. Codex with an API key skips cloud features, and Claude Code bills the Console account per token.
If you want each tool for different work, both $20 plans together cost $40 a month, less than either company's next tier.
Frequently asked questions
Is Codex cheaper than Claude Code?
Not at the entry price. ChatGPT Plus and Claude Pro both cost $20 a month, and Claude Pro drops to $17 billed annually. Through the API, Codex's GPT-6.1 Sol costs $2 and $10 per million input and output tokens, half the price of Claude Code's default Opus 5.5.
How much does Claude Code cost per month?
Claude Code comes with any paid Claude plan: Pro at $17 annual or $20 monthly, Max at $100, Ultra at $200, or a Team seat from $20. On the API instead, Anthropic says enterprise developers average $150 to $250 a month, about $13 per active day.
How much does Codex cost per month?
Codex comes with ChatGPT plans: Free and Go ($8) get GPT-6 Luna in the desktop app as it rolls out, Plus is $20, and Pro comes at $100, $200 or $500. ChatGPT Business is $20 per user per month billed annually. With an API key, you pay OpenAI's API rates instead.
Is Claude Max worth it over Pro for Claude Code?
Only if Pro's limits cost you work most weeks. Max is $100 for 5x Pro's usage every 5 hours and every week, which is the same $20 per Pro's worth of usage as monthly Pro. You pay for the bigger window, Fable access and $100 a month in API credits.
What's the difference between ChatGPT Pro $100 and $200 for Codex?
OpenAI's Codex pricing page lists both tiers with no five-hour limit and links to a help article for each tier's usage. Only Pro $500 adds Ultrafast mode. Check the help article and your usage dashboard before paying for the larger tier.
Do Claude Code and Codex share usage with chat?
Yes. Claude Code draws from the same pool as Claude chat on the web, desktop and mobile. Codex shares its usage with ChatGPT Work, and its local messages and cloud chats draw from one allowance. Heavy chat use leaves less for coding on both.
Can I use Claude Code or Codex with an API key instead of a plan?
Yes. Claude Code bills per token through a Console account, at Opus 5.5's $4 and $20 per million input and output tokens by default. Codex with an API key works in the CLI, SDK and IDE extension at API rates, without cloud features.
Make every plan go further
Both companies give the same advice for stretching a plan: keep the instructions file small. OpenAI's pricing page tells Codex users to reduce the size of their AGENTS.md and limit MCP servers, because both add context to every message. Anthropic's costs page suggests moving workflow-specific instructions out of CLAUDE.md into skills, which load only when invoked. Our guide to compound engineering for Claude Code shows how to build that file one fix at a time, so it stays short and every rule earns its tokens.
Stop guessing. Score your first workflow.
Score one workflow in about five minutes. Nine questions, a grade, and the specific gap holding it back - no call required.
Related Posts
Claude Code put account emails in User-Agent headers: what to update and check
Claude Code sometimes put signed-in users' account email into the User-Agent header of requests it wrote. Version 2.1.317 adds a check that asks first, the stable channel was still on 2.1.311, and the next release removes the email from the model's context. What to update, and what the check misses.
Claude Code parallel agents: the 5 ways and which one to use
How to run Claude Code agents in parallel: when to use subagents, agent view, agent teams, dynamic workflows or projects, how to give every session its own git worktree, and what parallel agents do to your token use.
Claude Code remote control: drive a local session from your phone
How to use Claude Code remote control: start a session with claude remote-control or /rc, connect from your phone or browser, run several sessions from one server, keep them alive, turn it on for every session, and fix the errors it prints.