“How much does Claude Code cost?” sounds like a one-line question. In practice it depends on how you sign in, which model you leave as the default, and how long your sessions run.
Anthropic also changed a lot this autumn: Opus 5.5 got cheaper, Claude Code switched Pro users to Opus by default, and on 1 November the $200 plan was renamed Ultra.
This guide puts every current price in one place, shows what a real coding session costs, and helps you pick the cheapest plan for the way you work.
- Is Claude Code free? No. The Free plan does not include it. The cheapest way in is Pro at $20/month ($17/month billed yearly).
- Plans with Claude Code: Pro $20, Max $100 (5x Pro), Ultra $200 (10x Pro weekly, no 5-hour limit once verified), Team Standard $25/seat, Team Premium $125/seat, Enterprise $20/seat + API-rate usage.
- Pay per token instead: Opus 5.5 $4/$20, Sonnet 5.5 $2/$10, Haiku 5.5 $0.10/$0.50 per million input/output tokens.
- Typical API spend: Anthropic’s own figure is about $13 per developer per active day, $150-250 a month, and under $30/day for 90% of users.
- Rule of thumb: under 1 hour of coding a day, Pro is enough. Daily full-time use, Max. Several agents in parallel all day, Ultra.
Claude Code Plans and Prices (November 2026)
Claude Code is included in every paid Claude plan. It shares one usage pool with Claude chat, so a long coding day also eats into your chat allowance.
| Plan | Monthly | Yearly | Claude Code usage | Extras |
|---|---|---|---|---|
| Free | $0 | n/a | Not included | Sonnet and Haiku in chat only |
| Pro | $20 | $17/mo ($200 up front) | Base allowance | Opus, Projects, Research, Claude in Chrome |
| Max | $100 | Monthly only | 5x Pro | Fable (up to half of weekly limit), $100/mo API credits, priority at peak |
| Ultra | $200 | Monthly only | 10x Pro weekly, no 5-hour limit when verified | Fable, $200/mo API credits |
| Team Standard | $25/seat | $20/seat | Above Pro | 2-150 seats, SSO, central billing |
| Team Premium | $125/seat | $100/seat | 5x Standard | Fable at 50% of weekly limit, pooled API credits |
| Enterprise | Sales | $20/seat + usage | Billed at API rates | SCIM, audit logs, HIPAA-ready option |
Note: prices are from Anthropic’s pricing page on 6 November 2026 and exclude tax. In Europe, Max and Ultra are €90 and €180 a month.
Is Claude Code Free?
There are three legal ways to run it without paying Anthropic a subscription:
- API credits: sign in with a Console account and pay only for the tokens you use. Light users can spend just a few dollars a month.
- Another model: point Claude Code at DeepSeek, OpenRouter or a local Ollama model. My Claude Code Router guide shows the exact settings.
- Local model:
ollama launch clauderuns it against a model on your own GPU for free, with weaker results than Opus or Sonnet.
How Usage Limits Work: 5-Hour and Weekly Windows
Subscriptions do not charge per token. Instead you get two allowances that refill on a timer.
- Session window: a rolling 5-hour window that starts with your first message. Ultra accounts that are verified no longer have this one.
- Weekly window: a cap across all models that resets at a fixed time for your account.
- Model limits: Opus and Fable are metered on top, so you can see “You’ve hit your Opus limit” while Sonnet still works.
How fast you burn through them depends on context size more than on how many messages you send. Claude Code resends the whole conversation with each request, so a one-line question in a session open all day still costs a lot.
Recent limit changes
| Date (2026) | Change |
|---|---|
| 13 May | Weekly limits raised 50% for Pro, Max, Team and seat-based Enterprise |
| 14 Sep | Weekly limit set permanently 25% above the pre-promotion level |
| 22 Sep | Opus 5.5 launched at $4/$20; Claude Code 2.1.280 made Opus the default on Pro and Team Standard; 5-hour limits raised |
| 1 Nov | Max 5x renamed Max, Max 20x renamed Ultra; Ultra loses the 5-hour limit after verification |
What Changed With Max and Ultra on 1 November
The two top plans kept their prices but got new names and slightly different allowances.
- Max ($100): the old Max 5x. Still 5x Pro usage, now with $100 a month of API credits.
- Ultra ($200): the old Max 20x. Weekly usage is listed as 10x Pro, not 20x, but the 5-hour window is gone once your account is verified. API credits rose from $100 to $200.
- Verification: new Ultra accounts keep a temporary 5-hour limit until Anthropic checks them for abuse. You get an email when it lifts. Existing subscribers needed to do nothing.
Tip: the lack of a 5-hour window matters most if you run long agent jobs overnight. If you work in short bursts, Max gives more value per dollar.
Claude Code With an API Key: Per-Token Prices
Sign in with a Console account (or set ANTHROPIC_API_KEY) and Claude Code bills every token at list price. API usage does not count against any subscription limit.
| Model (per million tokens) | Input | Output | Cache read | 5-min cache write | Batch (50% off) |
|---|---|---|---|---|---|
| Fable 5.1 | $10 | $50 | $0.25 | $12.50 | $5 / $25 |
| Opus 5.5 | $4 | $20 | $0.20 | $5 | $2 / $10 |
| Sonnet 5.5 | $2 | $10 | $0.10 | $2.50 | $1 / $5 |
| Haiku 5.5 (prompt up to 100k) | $0.10 | $0.50 | $0.01 | $0.125 | $0.05 / $0.25 |
| Haiku 5.5 (prompt over 100k) | $0.50 | $2.50 | $0.05 | $0.625 | $0.25 / $1.25 |
Three extras can change your bill:
- Fast mode on Opus 5.5 costs $8/$40, double the normal rate.
- US-only inference adds a 1.1x multiplier to everything.
- Thinking tokens are billed as output. Opus 5.5, Sonnet 5.5 and Haiku 5.5 always think, so lower
/effortinstead of trying to turn it off.
For the same table across OpenAI, Gemini, Grok and DeepSeek, see my AI API pricing comparison.
What a Real Session Costs (Worked Examples)
A typical one-hour coding session is mostly cache reads, because Claude Code re-sends the conversation every turn. These estimates use list prices and round numbers, not my own measurements.
| Session shape | Tokens | Sonnet 5.5 | Opus 5.5 |
|---|---|---|---|
| Small bug fix (20 min) | 500k cache read, 40k input, 15k output | about $0.28 | about $0.56 |
| Feature build (1 hour) | 2M cache read, 100k input, 60k output | about $1.00 | about $2.00 |
| All-day refactor with subagents | 15M cache read, 800k input, 400k output | about $7.10 | about $14.20 |
The last row lands close to Anthropic’s own average of $13 per active day. Agent teams use roughly 7x more tokens than a normal session, so multiply accordingly if you run them.
Warning: the first message after a long break misses the prompt cache and re-processes the whole context at the full input price. Long idle sessions are the most common cause of surprise bills.
Which Plan Should You Buy?
| How you code | Best option | Why |
|---|---|---|
| A few hours a week, hobby projects | Pro ($20) or API credits | Pro is cheaper than about $1/day of API use |
| 1-2 hours every working day | Pro, upgrade if you hit limits | Most users stay inside Pro’s window with Sonnet |
| Full-time developer, one session at a time | Max ($100) | API at $13/day is about $280 a month |
| Parallel agents, overnight jobs | Ultra ($200) | No 5-hour cut-off once verified |
| CI pipelines, scripts, product features | API key | Predictable per-token billing, batch discount |
| Company of 5+ developers | Team Standard or Premium | Central billing, SSO, per-seat allowance |
If you are still deciding between agents, my Cursor vs Windsurf vs Claude Code vs Cline comparison covers their prices side by side.
How to Turn On Usage Credits
Usage credits let Pro, Max and Ultra users keep working after a limit, billed at API rates. Here is how to set them up safely.
- Sign in with your subscription. Run
/loginin Claude Code and choose your claude.ai account. The command does not work with an API key. - Open the settings. Run
/usage-credits. It opens Settings, Usage on claude.ai in your browser. - Switch credits on. In the Usage credits section, turn them on and check your balance.
- Set a monthly spend limit. Without one,
/usageshows “Unlimited”. Pick a number you are happy to lose. - Check the row. Back in Claude Code,
/usagenow shows this month’s credit spend against your limit.
Note: on Team and Enterprise, members without billing access get a request sent to their admins instead.
How to Check What You Are Spending
/usage: plan usage bars, a breakdown by skill, subagent and MCP server, and recommendations. Pressdorwfor the last day or week.- Press
fin/usage(v2.1.305+) to let Claude Code apply the fixes it can, such as lowering effort. - Session block: API users see a dollar total per session. It resets on
/clear. - Status line: show cost and context use all the time. See my Claude Code commands cheat sheet.
- Console: API users get the authoritative bill at platform.claude.com, Usage.
--max-budget-usd: a CLI flag that stops a scripted run at a dollar cap.
10 Ways to Cut Claude Code Costs
- Use Sonnet by default. It costs half of Opus 5.5. Switch with
/modelonly for hard design work. - Run
/clearbetween tasks. Old context is re-read on every message. - Use
/compactwith instructions, such as “keep test output and code changes”. - Lower
/effortfor simple edits. Thinking is billed as output. - Give subagents Haiku with
model: haikufor searches and log reading. - Keep CLAUDE.md under 200 lines and move rare workflows into skills, which load only when needed.
- Disable unused MCP servers in
/mcp; prefer CLI tools likegh. - Write specific prompts. “Add validation to login() in auth.ts” beats “improve this code”.
- Use plan mode (Shift+Tab) on big tasks to avoid expensive rework.
- Filter noisy output with hooks, for example only passing failing test lines to Claude.
The biggest lever for API users is caching. My prompt caching guide explains the 5-minute and 1-hour cache lifetimes in detail.
Team and Enterprise Pricing
On Team and Enterprise, each member draws from a per-seat allowance with the same 5-hour and weekly windows, shared with chat and Cowork.
- Team Standard: $20/seat yearly or $25 monthly, 2 to 150 seats.
- Team Premium: $100/seat yearly or $125 monthly, 5x Standard usage. Give these seats to your heaviest Claude Code users.
- Enterprise: $20/seat plus usage at API rates, 20-seat minimum, billed yearly.
- Controls: admins see spend per user in org analytics and set spend limits per organisation, group or member.
For API-based teams, Anthropic suggests per-user rate limits that shrink as the team grows:
| Team size | Tokens per minute per user | Requests per minute per user |
|---|---|---|
| 1-5 | 200k-300k | 5-7 |
| 5-20 | 100k-150k | 2.5-3.5 |
| 20-50 | 50k-75k | 1.25-1.75 |
| 50-100 | 25k-35k | 0.62-0.87 |
| 100-500 | 15k-20k | 0.37-0.47 |
| 500+ | 10k-15k | 0.25-0.35 |
Limit Messages and What They Mean
“You’ve hit your session limit” or “weekly limit”
This is the shared plan window, so switching models does not help. Wait for the reset time shown, turn on usage credits, or (v2.1.234+) let Claude Code wait and continue the task automatically from /rate-limit-options.
“You’ve hit your Opus limit”
Only the Opus family is capped. Run /model and pick Sonnet to keep going until the Opus allowance resets.
“You’ve hit your monthly spend limit”
Your usage credits reached the cap you set. Raise it in Settings, Usage, or wait for the plan reset named in the message.
“/usage: rate limited until 3:14am”
Only the usage display is rate limited, not your work. Claude Code shows the last bars it loaded in the past hour; press r after the stated time.
429 or 529 errors with an API key
These are API rate or overload errors, not plan limits. My Claude API 429 and 529 fix guide covers retries and limits.
Usage climbs while you are idle
Scheduled /loop tasks, goal check-ins, agent teammates and messages from other sessions all send your full context. Stop loops you do not need and close teammates when done.
My Take
For most readers of this site, Pro at $20 is the right starting point. Set Sonnet as your default, clear context between tasks, and you will rarely hit the wall.
Move to Max the first week you hit limits more than twice. Ultra only makes sense if you run agents for hours unattended. Real cost control comes from habits, not the plan: an Opus session left open all day can burn more than a whole week of careful Sonnet use.
Want a full comparison of the chat side of these plans too? Read my Claude Free vs Pro vs Max vs Ultra guide.
FAQ
How much does Claude Code cost per month?
From $20 a month on Pro, $100 on Max and $200 on Ultra. With an API key you pay per token, which Anthropic says averages $150-250 a month for a full-time developer.
Is Claude Code included in Claude Pro?
Yes. Pro includes Claude Code and, since September 2026, uses Opus 5.5 as the default model. It shares one usage pool with Claude chat.
Can I use Claude Code on the free plan?
No. You need a paid plan or a Claude Console account with API credits. You can also connect it to a free local model through Ollama.
Is Max 20x still available?
It was renamed Ultra on 1 November 2026. The price stayed at $200, weekly usage is listed as 10x Pro, and verified accounts have no 5-hour limit.
Does API usage count against my subscription limits?
No. API key usage is billed separately in the Console and never touches your Pro, Max or Ultra allowance.
Is Opus or Sonnet cheaper in Claude Code?
Sonnet 5.5 costs exactly half of Opus 5.5 per token ($2/$10 against $4/$20). On a subscription, Opus also has its own extra limit.
How do I see how much I have used?
Run /usage in Claude Code. It shows plan bars, what used the most tokens, and suggested fixes you can apply with f.
What is the cheapest way to use Claude Code heavily?
Max at $100 if you code daily, since the same work through the API averages about $280 a month. For automation, the API’s 50% batch discount is cheapest.