Claude Code Pricing 2026: Pro vs Max vs API Costs

Updated on September 20, 2026 by Steven Jones

Claude Code doesn’t have a separate subscription fee. It is included with Claude Pro at $20 per month or $200 per year, Max 5x at $100 per month, and Max 20x at $200 per month. Developers can also pay for usage separately through usage credits or the Claude API.

The useful pricing decision is therefore not simply “how much does Claude Code cost?” It is whether your normal coding fits inside a subscription allowance, whether occasional overage is cheaper than upgrading, and whether your work is interactive or programmatic. This guide compares those routes, calculates the cash point at which an upgrade deserves testing, and covers a programmatic-usage billing detail that is easy to miss.

Pricing and billing details re-checked on 20 September 2026. US dollar prices are shown before applicable taxes or regional adjustments.

Claude Code pricing at a glance

Route	Price	How Claude Code is billed	Best fit
Claude Pro monthly	$20/month	Interactive Claude Code is included within shared plan limits	Short coding sessions and lighter repository work
Claude Pro annual	$200/year	Same Pro access, paid upfront	Developers expecting to keep Pro for a year
Max 5x	$100/month	5x Pro per-session usage capacity	Regular Claude Code use where Pro limits interrupt work
Max 20x	$200/month	20x Pro per-session usage capacity	Heavy daily interactive use
Plan plus usage credits	Plan fee plus metered usage	Included allowance first, then consumption-based billing	Occasional spikes beyond the subscription limit
Claude API	Pay per token and feature	Separate from the individual subscription allowance	Applications, CI and deliberately metered automation

Anthropic displays annual Pro as $17 per month because it rounds the $200 annual payment for display. The actual monthly equivalent is about $16.67. Paying annually saves $40 compared with twelve $20 monthly payments.

---

---

Pro and Max are usage allowances, not token bundles

Do not convert a Pro or Max subscription into a claimed number of API tokens. Anthropic doesn’t define consumer plans that way. Usage varies with the model, effort level, conversation length, repository context, tools and the amount of previous work Claude needs to carry into later turns.

Interactive usage across Claude and Claude Code shares plan limits. Max plans have a five-hour session limit plus weekly limits, and Anthropic can also apply limits to particular models or features. A long repository session can therefore affect the Claude capacity available elsewhere on the same account.

This is also why comparisons such as “$200 of Max equals thousands of dollars of API tokens” need caution. A fixed subscription allowance and metered API consumption are different products. Max can be extremely economical for sustained interactive coding, but it does not promise a fixed dollar value of compute.

Use your actual overage spend to decide whether Max is cheaper

The cleanest upgrade calculation uses money you actually spent, rather than estimating how many tokens a Max plan might represent.

Current plan	Next plan	Extra fixed cost to upgrade	Cash test
Pro monthly	Max 5x	$80/month	If you repeatedly spend more than $80 a month on extra usage and Max 5x would absorb most of it, test Max 5x
Max 5x	Max 20x	$100/month	If you repeatedly spend more than $100 a month on extra usage and Max 20x would absorb most of it, test Max 20x

For example, Pro plus $65 of extra usage costs $85 for that month. That is still $15 below Max 5x. Pro plus $110 of extra usage costs $130, which is $30 above Max 5x before tax. Likewise, Max 5x plus $140 of extra usage costs $240, compared with $200 for Max 20x.

Those are cash break-even checks, not claims that the higher subscription will reproduce the same amount of metered compute. Max capacity is not sold as an API-dollar equivalent. Check two things together: how much extra usage you paid for and whether subscription limits actually interrupted productive work. If both keep happening, the higher tier has a stronger case.

Claude Pro is still the lowest-cost paid route

Claude Pro costs $20 monthly or $200 annually and includes Claude Code. It remains the logical entry point for bug fixes, code explanations, contained refactors and occasional repository tasks.

A limit becomes economically relevant only when it starts changing how you work. One unusually large refactor is weak evidence for an $80 monthly upgrade. Repeated resets during ordinary development are much stronger evidence.

If you are still deciding whether your work needs repository access at all, our Claude AI for coding guide separates the jobs suited to Claude Chat, Claude Code and API access.

Max 5x vs Max 20x: the extra $100 needs a real workload

Max 5x costs $100 per month, and Max 20x costs $200. Both include Claude Code. The main difference for an individual developer is usage capacity, not a radically different coding feature set.

Work pattern	Plan to evaluate first	Reason to move up
A few short coding sessions each week	Pro	Move only if limits regularly interrupt useful work
Claude Code used most working days	Max 5x	Consider Max 20x if session or weekly limits remain a repeated bottleneck
Long repository agents and parallel interactive work	Max 20x	Higher allowance can reduce interruptions, but usage is still capped
Light most weeks, heavy around releases	Pro or Max 5x plus usage credits	Occasional metered overage may cost less than a permanent upgrade
Product or CI automation	API	Consumption can be measured and controlled directly

The annual difference between Max 5x and Max 20x is $1,200. Paying that pre-emptively makes little sense. Use Max 5x long enough to determine whether its limits create a repeatable problem.

Usage credits are metered usage, not a refill of your Max allowance

Pro and Max users can continue after reaching included limits by enabling usage credits. Once that happens, consumption is charged at standard metered rates rather than behaving like another block of subscription capacity.

This is where bills can accelerate. Repository rereads, large contexts, tool loops, subagents and long outputs can feel inexpensive while they are inside a subscription allowance, then become much more visible once every token is being priced directly.

Usage bundle	Discount	Price paid
$50 of usage	10%	$45
$250 of usage	20%	$200
$1,000 of usage	30%	$700

Bundles make occasional overage cheaper, but the upgrade calculation should use what you actually paid. If overage turns into a normal monthly expense, compare that expense with the $80 Pro-to-Max-5x or $100 Max-5x-to-Max-20x price gap.

The planned Agent SDK billing change is still paused

There is an important wrinkle for developers using Claude Code programmatically. Anthropic previously announced that Claude Agent SDK usage, claude -p and certain third-party Agent SDK applications would stop drawing from normal subscription limits and instead receive separate monthly credits.

That change did not take effect. As of 20 September 2026, Anthropic’s current Agent SDK billing note still says the change is paused. Agent SDK use claude -p and qualifying third-party app usage continue to draw from the subscription’s usage limits for now, and the proposed separate monthly credit is not available.

This affects cost planning for automated coding workflows. Do not build a budget around the previously proposed $20 Pro, $100 Max 5x or $200 Max 20x monthly Agent SDK credits unless Anthropic activates a replacement scheme later.

Claude API pricing for coding workloads

API billing remains separate from the individual Pro and Max subscriptions. The main current model prices relevant to common coding workloads are:

Model	Input per million tokens	Output per million tokens	5-minute cache write	Cache read
Claude Sonnet 5	$2	$10	$2.50	$0.20
Claude Opus 5	$5	$25	$6.25	$0.50
Claude Haiku 4.5	$1	$5	$1.25	$0.10

Prompt caching is particularly relevant to repository agents because stable context may be reused repeatedly. On Sonnet 5, one million cache-read tokens cost $0.20 rather than the $2 base input rate. New input, cache writes, output, and tool-driven model calls still contribute separately, so multiplying every token in a long coding session by the base input price can substantially overstate a well-cached workload.

Worked API costs show how quickly automation can become expensive

These examples use uncached list pricing so the arithmetic remains reproducible. They are illustrative workloads, not measured Claude Code sessions.

Illustrative workload	Token pattern	Sonnet 5	Opus 5
Small repository fix	300k input + 30k output	$0.90	$2.25
Medium feature session	2M input + 200k output	$6	$15
Heavy agent session	8M input + 600k output	$22	$55
20 medium sessions	40M input + 4M output	$120	$300
20 heavy sessions	160M input + 12M output	$440	$1,100

A $200 Max 20x subscription can therefore look dramatically cheaper than a heavy month priced at API rates. Do not turn that comparison into a claimed Max token allowance. The subscription gives access under usage controls; the API charges for measured consumption.

An API key can make Claude Code bill the wrong account

One configuration mistake can bypass the subscription entirely. If an ANTHROPIC_API_KEY environment variable is configured, Claude Code can authenticate with that API key instead of the Pro or Max subscription, causing API charges.

Run /status when you are unsure which account or authentication route is active. For metered sessions, /usage or its /cost alias can help expose session consumption. This check is particularly useful after moving machines, changing shell profiles or working in a project that injects environment variables automatically.

Which Claude Code pricing route makes sense?

Pro is the lowest-risk starting point for interactive Claude Code. Pay annually only if you are comfortable committing for the year.

Max 5x deserves testing once Pro limits repeatedly interrupt normal work or your real monthly overage approaches the $80 price difference.

Max 20x is easier to justify after Max 5x has demonstrated a capacity problem, particularly if you are also spending more than $100 a month on extra usage.

API billing suits workloads where per-request accounting, automation, explicit model selection and programmatic spend controls matter more than a fixed subscription.

Subscription plus usage credits works best when heavy demand is occasional. Keep the base plan until the overage history shows that a higher fixed tier is cheaper.

If price is making you consider another coding agent, compare the AI coding tools developers keep recommending before adding another subscription.

Your search, your sources

Make DIY AI a preferred source

See more of our reporting in Google Top Stories, AI Overviews and AI Mode.

Add DIY AI on Google

How to reduce Claude Code cost before upgrading

• Separate tasks cleanly. Do not keep irrelevant history attached to a new repository problem.
• Control context deliberately. Large directories and unnecessary files consume capacity even when they add little useful information.
• Use cheaper models when the task permits. Reserve more expensive reasoning for work that benefits from it.
• Check authentication. Confirm that subscription use has not silently become API-key billing.
• Record real overage for a month. Use actual spending and interruptions to decide the next plan.
• Put a cap on usage credits. A fixed subscription is predictable; metered continuation is not.

Verdict: upgrade from measured cost, not theoretical token value

For an individual developer, Claude Pro remains the cheapest sensible paid starting point at $20 per month or $200 per year. Max 5x becomes more compelling once Pro repeatedly blocks useful work or your actual overage starts approaching the $80 monthly upgrade difference. Max 20x should solve a Max 5x capacity problem you have already observed.

API billing is easier to measure but can become expensive during context-heavy agent workflows. Usage credits help with spikes, but monitor them as real metered spend rather than another slice of your subscription allowance.

You can access Claude Code through the paid Claude plans. If you want to test the product before paying, our Claude Code free-tier guide covers the current no-cost access options and their limitations.

Claude Code pricing FAQs

How much does Claude Code cost per month?

Claude Code has no separate individual subscription fee. It is included with Claude Pro at $20 per month, Max 5x at $100 per month and Max 20x at $200 per month. Pro is also available for $200 billed annually.

Is Claude Code included in Claude Pro?

Yes. Claude Code is included with Claude Pro. Interactive usage shares the same broader subscription limits as other Claude activity on the account.

Does Claude Pro include API credits?

No general Claude API allowance is included for applications using a Console API key. Pro users can separately enable usage credits to continue after subscription limits, but that continuation is metered.

Does Claude Code have a fixed token limit on Pro or Max?

Anthropic does not publish the individual subscriptions as fixed token bundles. Consumption depends on factors including model, effort, context length, conversation history and tool use. Plans have session and weekly usage controls.

When is Max 5x cheaper than Pro plus usage credits?

Max 5x costs $80 more per month than monthly Pro. If you repeatedly spend more than $80 on Pro overage and the additional Max 5x capacity would prevent most of that spend, Max 5x deserves testing. The two plans are not token-for-token equivalents, so use actual bills and limit interruptions rather than theoretical API value.

Does claude -p use a separate Agent SDK allowance?

Not currently. Anthropic announced a separate monthly Agent SDK credit system but paused the change before it took effect. As of 20 September 2026, claude -p and Agent SDK usage continues to draw from subscription usage limits unless you deliberately authenticate and pay through another billing route.