Claude Code Pricing
Pick a subscription for a recurring allowance and the Claude apps, or API billing when you want metered usage, explicit model costs and per-project spend control.
On this page
- Claude Code pricing at a glance
- What a Claude subscription pays for
- What Claude Code costs with the API
- Claude Code API cost examples
- Subscription vs API: which is cheaper?
- Calculate cost per accepted task
- Control API spend without breaking legitimate work
- Common pricing mistakes
- Comparing Pro with one API token rate
- Assuming Max is unlimited
- Treating the context window as included usage
- Using /usage list-price estimates as the gateway invoice
- Ignoring cache writes and thinking tokens
- Frequently asked questions
- How much does Claude Code cost per month?
- Is Claude Code included in Claude Pro?
- Is the Claude Code API cheaper than a subscription?
- What happens when I reach a Claude Code subscription limit?
- Can I use a subscription and an API key together?
- Does /usage show my real OmniaKey cost?
- Sources
Updated October 24, 2026: new Team seat prices, Max API credits and Claude 5.5 API prices.
Claude Code costs $17 a month on Pro if you pay for a year ($200 up front), $20 a month on monthly billing, $100 a month on Max 5x, and $2 per million input tokens on Sonnet 5.5 when you pay the API instead. Anthropic's pricing page (checked October 24, 2026) lists those numbers, and below they turn into the cost of one real coding task.
The short answer:
- pick Pro at $17–20 a month for regular interactive work: Claude Code, the Claude apps and one shared allowance;
- pick Max from $100 when Pro limits interrupt you, or the API when usage is occasional, automated or billed to a project;
- compare cost per accepted task, not a monthly price against one token rate.
Every Claude plan, not only Claude Code, is compared in Claude pricing.
Cowork is a separate paid-plan workflow and has no published API-key path. If you are choosing between Claude agents before comparing cost, start with Claude Code vs Cowork.
A subscription is not a bucket of API credits. Anthropic publishes no fixed token amount for Pro or Max. Limits depend on conversation length, model, features and usage patterns, and can include rolling session, weekly, model or feature limits.
Prices and product facts were checked October 24, 2026. Every amount below is in USD before local taxes. Plans, limits and model rates change, so confirm on the linked Anthropic pages and in the live OmniaKey catalog before you buy.
Claude Code pricing at a glance
Anthropic sells Free, Pro and Max to individuals. Team and Enterprise add seats, admin controls and pooled usage on top of the same Claude Code client.
| Plan | Price | Claude Code | Included |
|---|---|---|---|
| Free | $0 | No | Sonnet and Haiku |
| Pro | $17/mo yearly or $20/mo | Yes | Opus, Sonnet, Haiku; Fable via usage credits |
| Max 5x | $100/mo | Yes | 5× Pro usage, $100 API credits |
| Max 20x | $200/mo | Yes | 20× Pro usage, $200 API credits |
| Team Standard | $20/seat yearly, $25 monthly | Yes | Up to $500/mo pooled API credits |
| Team Premium | $100/seat yearly, $125 monthly | Yes | 5× Standard usage, Fable up to 50% |
| Enterprise | $20/seat/mo yearly | Yes | Usage at API rates |
The yearly Pro price is $200 charged once, which works out to $17 a month. Max costs $100 for the 5x plan or $200 for 20x, and both let Fable use up to 50% of your weekly limits. Team runs from 2 to 150 seats: Standard seats cost $20 a seat a month on annual billing or $25 monthly, Premium seats cost $100 or $125 and give five times the Standard usage plus Fable up to 50% of weekly limits. Enterprise is $20 per seat per month billed annually with a 20-seat minimum, plus usage at API rates. Free costs $0 and does not include Claude Code.
Those are headline prices. Taxes, annual prepayment and optional usage credits change the invoice, and a bigger allowance is not a guaranteed number of prompts. One short question after a long repository session can carry more context than several clean-session questions.
What a Claude subscription pays for
Pro and Max combine Claude Code with the rest of Claude. The recurring payment covers an included allowance rather than a per-request invoice, so your monthly cash cost is easy to predict until you hit a limit, while your cost per token stays hard to calculate.
Subscription usage is drawn from one pool by:
- Claude Code in the terminal or the IDE;
- Claude on the web, desktop and mobile;
- long conversations and repeated context;
- the selected model and reasoning effort;
- subagents, agent teams, MCP tools and other model-backed features.
The /usage screen makes that visible. API accounts see session tokens and a local estimate at Anthropic's list rates. Subscribers see plan bars, recent attribution and the behaviours eating their allowance.
When a paid subscriber reaches a plan limit, Anthropic can offer usage credits at standard API rates. The account stays a subscription, but work beyond the allowance becomes metered, and you can cap that spend. The Claude Code docs say /usage-credits appears after signing in with a Claude.ai subscription; it is a subscription feature, not an API-key login.
What Claude Code costs with the API
API billing separates the Claude Code client from the subscription. The CLI can sign in through Claude Console, a supported cloud provider or an authorised gateway, and each has its own rate card and spend controls.
Anthropic's API prices per million tokens (checked October 24, 2026):
| Model | Input | Cache read | Output |
|---|---|---|---|
| Fable 5.1 | $10 | $0.25 | $50 |
| Opus 5.5 | $4 | $0.20 | $20 |
| Sonnet 5.5 | $2 | $0.10 | $10 |
| Haiku 5.5 | from $0.10 | $0.01 | from $0.50 |
Haiku 5.5 charges $0.10 per million input tokens for prompts up to 100K tokens and $0.50 above that, with output at $0.50 and $2.50 per million in the same two bands. Fable 5.1 keeps Fable's $10 and $50 rates and bills cache reads at $0.25.
Opus 5 and Opus 4.8 still list at $5 per million input and $25 per million output, so Opus 5.5 is 20% cheaper on both and 60% cheaper on cache reads ($0.20 against $0.50). A Sonnet 5.5 session is half the price of an Opus 5.5 session on input and output, which is why opusplan — Opus for planning, Sonnet for execution — is a sensible default for long jobs.
Read the live pages for Claude Fable 5.1, Claude Opus 5.5, Claude Sonnet 5.5 and Claude Haiku 5.5 before you choose. Cache writes cost more than reads: $12.50 or $20 per million for Fable 5.1, $5 or $8 for Opus 5.5 and $2.50 or $4 for Sonnet 5.5, depending on whether you keep the cache for five minutes or an hour.
For Fable's official-versus-platform arithmetic and the steps for API access, read the Claude Fable 5 pricing guide.
Through OmniaKey, the same models run behind one key and one balance. Use the live model catalog for the current gateway quote, then reconcile it against the usage dashboard.
Claude Code API cost examples
Start with a token mix, not a guessed number of prompts:
request cost =
uncached input × input rate
+ cached input × cache-read rate
+ cache writes × cache-write rate
+ generated and thinking tokens × output rate
+ provider tool charges
These examples use Sonnet 5.5 at $2 per million input and $10 per million output, no cache, no paid provider tools and no retries. They show scale; they do not predict how many files or turns your task needs.
| Task | Tokens | Cost |
|---|---|---|
| Small fix | 250K in + 20K out | $0.70 |
| Multi-file feature | 1M in + 100K out | $3.00 |
| Long investigation | 5M in + 500K out | $15.00 |
The arithmetic for all three, at Sonnet 5.5 list prices:
small: 0.25 × $2.00 + 0.02 × $10.00 = $0.50 + $0.20 = $0.70
feature: 1.00 × $2.00 + 0.10 × $10.00 = $2.00 + $1.00 = $3.00
long: 5.00 × $2.00 + 0.50 × $10.00 = $10.00 + $5.00 = $15.00
Real Claude Code sessions usually have a large cache-read component, because history, project instructions, tool definitions and previously read files get reused. Cache reads cost a tenth of the input rate on Sonnet 5.5 and Opus 5.5, so a long session with a stable prompt prefix is cheaper than the raw input numbers suggest. Output-billed thinking tokens can still run higher than the visible answer. Measure the finished task in /usage and reconcile it against the provider's usage record.
Subscription vs API: which is cheaper?
At list prices the break-even is easy to compute, even though Anthropic does not sell Pro or Max as a fixed token bundle. A $20 Pro month buys the same tokens as about six multi-file tasks at $3.00 each ($20 ÷ $3.00 = 6.7), or 28 small fixes at $0.70 ($20 ÷ $0.70 = 28.6). If you finish more than six multi-file tasks in a month, Pro at $20 costs less than the same work on the API; on the annual price of $17 the line sits at five tasks ($17 ÷ $3.00 = 5.7). Max 5x at $100 covers 33 such tasks at list price and Max 20x at $200 covers 66.
| Your usage | Start with | Why |
|---|---|---|
| Regular interactive work | Pro | One payment covers Claude Code and the Claude apps |
| Pro limits interrupt you daily | Max, or Pro plus capped credits | More allowance before you wait |
| Occasional coding sessions | API or prepaid gateway | Pay only when requests run |
| CI, scripts, scheduled jobs | API or gateway | Programmatic keys, per-project attribution, hard caps |
| Several coding tools and model families | Gateway | One balance for compatible tools |
| Web Claude Code or plan-only features | Subscription | API keys do not unlock them |
| A team needing identity and admin | Team or Enterprise | Seats include organisation controls |
For a light user, API billing avoids paying during quiet months. For a heavy interactive user, Pro or Max usually delivers far more work than the token-equivalent price suggests, because the allowance is sized for daily use rather than for one task. For automation, a metered key is easier to attribute and cap even when a subscription already exists for human work.
Using both is normal. Keep the subscription for interactive and app work, then give automation a separate API key with its own budget. Do not share a personal subscription login with a server and do not treat a consumer allowance as an automation pool.
Calculate cost per accepted task
Token price answers only part of the buying question. A cheap run that fails twice costs more than an expensive run that passes once. Track a unit your team values:
accepted-task cost =
model and tool charges
+ rerun charges
+ reviewer correction time
+ remediation cost for escaped mistakes
Use the same task class and the same checks when you compare options, and record:
- selected model and effort;
- uncached input, cache reads, cache writes, output and thinking;
- number of model turns and retries;
- whether tests and required checks passed;
- human correction time;
- the final provider charge.
Anthropic reports that enterprise API deployments average around $13 per developer per active day and $150–250 per developer per month, with 90% staying below $30 per active day. Those are broad deployment observations, not a forecast for your repository or a Pro-plan equivalent. A small pilot remains the honest way to budget.
The Claude Code token-usage guide covers context, tools, effort and reruns in more detail.
For a direct rate-card comparison across Claude and GPT tiers, including caching, Batch and long context, use the Claude API vs OpenAI API pricing guide.
Control API spend without breaking legitimate work
Set the cap from observed task cost. Start with one scoped key and one representative task, then put the ceiling above normal variance.
With OmniaKey:
- create separate keys for local work, CI and scheduled automation;
- give each key a limit that matches its job;
- inspect the requested model, tokens, latency and actual cost per call;
- investigate loops or unexpected models before raising a limit;
- revoke one job without interrupting the others.
The balance and usage guide covers the exact controls. The pay-as-you-go Claude Code guide walks through the legitimate login options and setup.
Common pricing mistakes
Comparing Pro with one API token rate
Pro includes several Claude products and a variable allowance; the API measures model consumption. Compare the work each option completes over a real month, not two numbers from different tables.
Assuming Max is unlimited
Max provides more usage than Pro, never unlimited use. Rolling session, weekly, model and feature limits still apply.
Treating the context window as included usage
A one-million-token context window is capacity. On the API, every token you place in it is billed as input or cache.
Using /usage list-price estimates as the gateway invoice
Claude Code calculates its dollar estimate locally at Anthropic's list rates. Gateway promotions differ. Reconcile against the provider dashboard.
Ignoring cache writes and thinking tokens
Cache reads are cheap, cache writes are not, and extended thinking is billed as output on API accounts even when the final answer is short.
Frequently asked questions
How much does Claude Code cost per month?
Free costs $0 and does not include Claude Code. Pro is $17 a month billed annually ($200 up front) or $20 monthly. Max starts at $100 a month, Team Standard at $20 a seat a month billed annually, Enterprise at $20 a seat a month with a 20-seat minimum. API-billed Claude Code has no monthly price: the bill follows model tokens, cache behaviour, tools and retries.
Is Claude Code included in Claude Pro?
Yes. Pro includes Claude Code, and its usage shares one allowance with the rest of Claude rather than a separate API-token balance. Pro also gives Opus, Sonnet and Haiku, with Fable available through paid usage credits.
Is the Claude Code API cheaper than a subscription?
Often yes for occasional or tightly scoped work, because nothing is billed between sessions. At Sonnet 5.5 list prices, $20 of API credit equals about six multi-file tasks or 28 small fixes, so Pro wins as soon as you code most days. Max 5x at $100 covers roughly 33 multi-file tasks at the same rates.
What happens when I reach a Claude Code subscription limit?
Wait for the displayed reset, move to a bigger plan, or turn on capped usage credits on an eligible paid plan. Switching models does not reset a shared session or weekly limit, though it can help after a model-specific cap.
Can I use a subscription and an API key together?
Yes. Keep the credentials explicit: subscription OAuth for interactive work, a separate capped API key for CI. Check the active login with /status and the billing dashboard.
Does /usage show my real OmniaKey cost?
It shows token counts and a local estimate based on Anthropic's list prices. Use the OmniaKey usage dashboard for the actual gateway charge.
Sources
Related articles
Claude Pricing 2026: Every Plan, Usage Limit and Model Included
Every Claude plan, its price and its usage limit in one place, plus the API rates and the plan that actually includes Fable.
Claude Code API Without a Subscription
Claude Code's local terminal can use API billing instead of a monthly Claude plan. You still pay for model usage; the difference is how authentication and billing work.
Best model for Claude Code
Start daily coding with Sonnet 5.5 at $2 per million input tokens, move complex multi-file work to Opus 5.5 at $4, keep Fable 5.2 at $8 for the hardest long sessions, and hand quick edits and subagents to Haiku 5.5.