Claude Code Pricing 2026: Pro vs Max vs API Costs

Updated on October 17, 2026 by Steven Jones

Claude Code pricing starts at $20 per month with Claude Pro; Max 5x costs $100 per month, and Max 20x costs $200 per month. There is no separate individual Claude Code subscription fee on these plans. Developers can also use metered usage credits or pay for Claude through the API.

The right route depends on how frequently your coding sessions hit usage limits and whether you also run scripts or applications through the Claude API. This guide compares current plan allowances, shows a cost example for every billing route and explains the new Max API credits without confusing them with extra Claude Code usage.

Pricing checked: 16 October 2026. Prices are in US dollars before applicable tax and regional differences. Examples are illustrative, not measured Claude Code sessions.

Claude Code pricing at a glance

Billing route	Price	What you get	Best fit
Claude Pro monthly	$20/month	Claude Code within shared subscription limits; no monthly API credit	Occasional repository work
Claude Pro annual	$200/year upfront	Same Pro usage, equivalent to $16.67/month	Committed year-round users
Max 5x	$100/month	5x Pro per-session capacity plus $100/month in separate, claimable API credits	Frequent interactive coding
Max 20x	$200/month	20x Pro per-session capacity plus $200/month in separate, claimable API credits	Very heavy interactive coding
Pro or Max plus usage credits	Plan fee plus metered overage	Continue interactive work beyond included limits	Occasional busy periods
Claude Console / API	Pay per token and feature	Metered access for applications, agents and scripts; eligible Max API credits may offset charges	Automation and product integrations

Max multipliers describe per-session allowances, not a promise of five or twenty times as many tokens throughout the month. Both Max subscriptions are monthly; Pro is also available annually for $200, saving $40 compared with twelve monthly payments.

---

---

What would each Claude Code billing route actually cost?

Route	Illustrative month or purchase	Cash cost
Pro monthly	Several coding sessions, all inside included limits	$20
Pro annual	One-year commitment, with no extra usage purchased	$200 upfront, or $16.67/month equivalent
Max 5x	Interactive coding within limits, plus $60 of eligible Console API work covered by claimed credits	$100
Max 20x	Interactive coding within limits, plus $220 of eligible API work; $200 covered by claimed credits	$220, assuming the remaining $20 is paid separately
Pro plus usage bundle	$20 Pro plus a $50 face-value extra-usage bundle purchased for $45	$65 upfront
API only	2 million uncached Sonnet 5.5 input tokens plus 200,000 output tokens	$6, excluding other billable features

The Max examples assume the subscriber is eligible, has claimed the monthly credits into a linked Claude Console organisation and is using them on qualifying API work. Unused monthly API credit expires, and it does not reduce the subscription invoice. Buying a usage bundle also doesn’t guarantee you’ll use the entire balance that month.

Pro and Max usage limits are not fixed token bundles

Claude Pro, Max 5x and Max 20x have rolling five-hour session limits and weekly limits. Usage is shared across Claude’s other interfaces and Claude Code, including supported IDE integrations. Repository size, model choice, effort level, conversation history, tool calls, and parallel agents all affect how quickly you use the allowance.

Max 5x offers five times Pro’s per-session usage, while Max 20x offers twenty times. Weekly caps still apply, and Anthropic can place additional controls on particular models or features. A developer can therefore hit a weekly limit without consistently exhausting every five-hour window. There is no reliable conversion from a $200 subscription to a specified dollar value of API tokens.

One repeatable pattern in developer experiences is that ordinary single-agent work fits comfortably within a Max plan, while a burst of parallel subagents consumes the allowance much faster. Track the work that triggers each reset, not just the number of days you opened Claude Code.

There is another model-specific exception: Fable 5 and 5.1 use pay-as-you-go usage credits on Pro rather than Pro’s included allowance. On Max, Fable usage is included but subject to a separate cap within the existing weekly limits. Choosing a premium model can therefore alter the cost calculation even without changing subscription tiers.

Max now includes API credits, but they will not refill Claude Code

Anthropic added monthly Claude Platform API credits to Max plans on 7 October 2026: $100 on Max 5x and $200 on Max 20x. These differ from the usage credits you buy to continue interactive Claude Code after you hit a limit.

Balance	Can pay for	Cannot pay for
Max monthly API credit	Qualifying Claude Console API calls, Playground, Managed Agents and API-key-authenticated Agent SDK use	Interactive Claude Code, Claude apps or their extra usage
Purchased usage credits/bundles	Metered continuation after the included Claude or Claude Code allowance is reached	A larger included five-hour or weekly subscription limit
Purchased Console API credits	Metered API calls and eligible Claude Code work deliberately billed through Console	A guarantee that subscription limits will increase

New Max subscribers can claim the API benefit after seven days on an eligible plan. In Claude’s Settings > Billing, link a Claude Console organisation with appropriate permissions. Credits are allocated to that organisation each billing cycle, used before purchased API credits and expire if unused. You can link only one organisation without support assistance. Anthropic explains the coverage and eligibility in its monthly API credit guidance.

This changes one upgrade calculation. Suppose Pro meets your interactive needs, but you also spend $90 every month on eligible Claude API requests. Pro plus that API bill would cost $110. Max 5x could cover the $90 from its claimed $100 API credit, making the total $100. That is a genuine $10 difference, provided the credit is available and the requests qualify. For somebody who never uses the API, the new Max credit has no cash value.

When does upgrading from Pro to Max make financial sense?

Without qualifying API usage, compare your actual extra-usage spending with the subscription price gap. Moving from monthly Pro to Max 5x costs another $80 a month; moving from Max 5x to Max 20x costs another $100. These are cash comparisons, not equivalent amounts of model computation.

Example: Pro plus $65 of metered extra usage totals $85, still $15 below Max 5x. Pro plus $110 totals $130, or $30 above Max 5x. The higher tier is worth trying only if its included capacity is likely to eliminate most of that overage. A larger allowance can still hit a weekly cap, particularly in context-heavy or multi-agent sessions.

The API-credit calculation is separate. If you consistently make $120 of qualifying API calls each month, Pro plus API costs $140, while Max 5x with a fully claimed and used $100 credit costs $120. Max 20x’s extra $100 credit can partly offset its $100 higher fee than Max 5x, but it isn’t automatically a saving: at $150 of API demand, Max 5x would cost $150 overall, and Max 20x would still cost $200. Upgrade to 20x primarily when 5x’s interactive limits are a demonstrated obstacle.

Usage bundles are a middle route for occasional overage

Pro and Max subscribers can enable metered usage credits to keep working after five-hour or weekly limits. Anthropic also offers discounted prepaid bundles to continue.

Usage value	Discount	Purchase price
$50	10%	$45
$250	20%	$200
$1,000	30%	$700

These bundles are for usage beyond subscription limits. They are not the free monthly Max API benefit, and they do not turn a Max 5x allowance into Max 20x. When comparing plans, use the discounted amount actually paid, not the face value of the bundle. A developer with one unusually busy release week may be better served by a smaller bundle than by paying an extra $80 every month.

Claude API pricing for coding workloads in October 2026

For Console-based coding agents and applications, model pricing is metered. Anthropic’s newer Opus 5.5, Sonnet 5.5, and Haiku 5.5 rates replace the older model selection in the previous version of this article.

Model	Input / 1M tokens	Output / 1M tokens	5-minute cache write / 1M	Cache read / 1M
Claude Sonnet 5.5	$2	$10	$2.50	$0.10
Claude Opus 5.5	$4	$20	$5	$0.20
Claude Haiku 5.5, prompts up to 100k tokens	$0.10	$0.50	$0.125	$0.01
Claude Haiku 5.5, prompts over 100k tokens	$0.50	$2.50	$0.625	$0.05

Haiku’s higher rate applies to each request whose prompt exceeds 100,000 tokens, not when aggregate monthly input crosses that figure. Sonnet 5.5 and Opus 5.5 have standard rates across their supported long-context windows. Real totals may also include cache writes, tools, fast-mode premiums or geographical routing charges where selected.

Worked API costs: small fixes to sustained agent work

The calculations below assume all tokens are uncached and billed at standard Sonnet 5.5 or Opus 5.5 rates. They show what consumption costs, not how many tokens a Pro or Max subscription includes.

Illustrative workload	Input + output tokens	Sonnet 5.5	Opus 5.5
Small repository fix	300k + 30k	$0.90	$1.80
Medium feature session	2M + 200k	$6	$12
Heavy agent session	8M + 600k	$22	$44
20 medium sessions	40M + 4M	$120	$240
20 heavy sessions	160M + 12M	$440	$880

Cost formula: (input tokens / 1,000,000 x input rate) + (output tokens / 1,000,000 x output rate). The medium Sonnet example is (2 x $2) + (0.2 x $10) = $6.

Prompt caching can substantially cut repeat-context charges. On Sonnet 5.5, one million input tokens served as cache reads costs $0.10 rather than the $2 standard input rate. Do not apply that discount to newly read files, new output, cache writes or every call in a tool loop. Changing the model can also change output quality, retry count and total tokens consumed, so identical token totals are a costing exercise rather than an equal-quality comparison.

Does claude -p or the Agent SDK use your subscription or API credits?

Anthropic paused its earlier proposal for separate Agent SDK subscription credits in June. The October update did not make all programmatic work API-only. Qualifying claude -p and Agent SDK work authenticated through your Claude subscription still uses subscription allowances.

There is now a second valid route: run claude -p or an Agent SDK application yourself using an API key from the Console organisation linked to an eligible Max plan. That qualifying Agent SDK usage can draw on its monthly API credit. Interactive Claude Code, IDE extensions and the Claude Code GitHub Action do not qualify merely because a command runs non-interactively.

For CI pipelines, keep the authentication and billing route explicit. A scheduled agent using Console credentials has measurable token costs; a plan-authenticated session draws on variable shared limits. Treat them as separate budgets, even when the same developer uses both.

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

Check the authentication route before blaming your subscription

An ANTHROPIC_API_KEY environment variable can cause Claude Code to authenticate through Console rather than the Pro or Max subscription. In non-interactive mode, the variable is used when present; in interactive mode, Claude Code can ask for approval before overriding the subscription. Either route can produce API billing when you expected included usage.

• Run /status in Claude Code to inspect your active account and limits before a long session.
• Use /usage (or /cost) to review session usage and the available plan information.
• If your shell is intentionally meant to use the subscription, remove an unwanted API-key setting and sign back in through the Claude plan. On Unix-style shells, unset ANTHROPIC_API_KEY clears it for the current shell.
• Review Console auto-reload and set sensible controls before running unattended API agents.

For a comparison of what repository agents are actually good at, rather than their subscription prices, see our Claude AI for coding assessment. Developers considering a competing coding agent can also read Claude Code vs OpenAI Codex.

Verdict: choose the cheapest route that removes a real bottleneck

Start with Claude Pro if you mainly work interactively and have not demonstrated that its limits interrupt normal development. Annual Pro becomes cheaper than monthly if you use the service for at least eleven months of the year. Test Max 5x when Pro’s resets and metered overage create a recurring cost. Reserve Max 20x for an observed 5x capacity problem or a combined coding and qualifying API workload that supports the additional fee.

The October API credit benefit makes Max more attractive to developers already paying for Claude-powered applications, but it is not an interactive Claude Code top-up. If you only need automation, calculate the API bill first. If you need occasional extra sessions, check actual discounted usage-bundle spend before committing to a higher subscription.

Claude Code pricing FAQs

How much does Claude Code cost a month?

Claude Code is included in individual Claude Pro at $20 monthly, Max 5x at $100 monthly and Max 20x at $200 monthly. Pro also costs $200 annually if paid upfront. Console-based API use and extra usage can add charges.

Is Claude Code free with Claude Pro?

It is included in the Pro subscription with no separate Claude Code fee, but it’s not unlimited. The allowance is shared with other Claude activity and subject to five-hour and weekly limits.

Does Claude Pro include API credits?

No. The recurring Console API credit benefit introduced in October 2026 applies to eligible Max and Team plans, not Pro. Pro subscribers can separately purchase usage credits for interactive overage or Console API credits for their applications.

Can I use Max’s $100 or $200 API credits when Claude Code hits a limit?

No. Monthly Max API credits cover eligible Console API and Agent SDK usage, not interactive Claude Code or its extra usage. You need purchased usage credits, a higher tier or a limit reset to continue interactive work.

Are Max 5x and Max 20x token allowances?

No. The multipliers describe relative per-session usage capacity compared with Pro. Weekly limits, model-specific rules and task complexity still affect what you can complete.

Is Claude Code cheaper through the API?

For a small number of measurable calls, it can be: the illustrative medium Sonnet 5.5 session above costs $6 at uncached API rates. Heavy repeated repository agents can cost much more than a fixed interactive subscription. Compare the total cost of finished tasks, including retries and time lost to limits, rather than treating the two routes as identical.