Claude Pricing Explained: Every Plan in 2026

What does Claude actually cost?

Claude pricing splits into two billing systems that confuse almost every new user. There are subscriptions for interactive use (Free, Pro, Max, Team, Enterprise) and there is the API for programmatic use (pay per token). They run on separate accounts, separate usage pools, and separate terms of service.

This guide covers every plan available in 2026, what each one actually includes, and which user type each plan fits. If you only want the Claude Code product specifically, our dedicated Claude Code pricing breakdown goes deeper on the developer plans, and if you are deciding between ecosystems, Claude vs ChatGPT pricing compares the two head to head for developers. Otherwise, read on for the full Anthropic ecosystem.

The two billing systems

The most important thing to understand before picking a plan:

Subscriptions charge a flat monthly fee. They cover Claude on the web, the desktop app, the mobile apps, and Claude Code in the terminal. All four interfaces draw from the same usage pool. Subscriptions explicitly forbid automated and scripted usage in their terms of service.

The API charges per input and output token consumed. There is no monthly fee, only what you actually use. The API is required for any automated, scripted, batched, or production workload. It also unlocks higher concurrency than subscription tiers allow.

A single user typically picks one or the other. A solo developer using Claude Code interactively belongs on a subscription. A team running automated pipelines or building Claude into a product belongs on the API. Picking the wrong system means either overpaying for limits you do not use or hitting rate caps that should not apply to your workload.

Free tier

Cost: $0. Available without signing up for billing.

The Free tier gives you Claude on the web, on iOS and Android, on the desktop app, and in projects, with limits sized for casual chat use. You can use any model (Sonnet, Opus, Haiku), generate code, analyse images, search the web, and create files. Memory is included. The cap is daily message volume, not access to capabilities.

Free is not viable for sustained development work. You will hit the daily cap within an hour of any real coding session. It is genuinely useful for trying Claude before paying, for occasional writing assistance, and for teams evaluating whether a Pro subscription is worth the upgrade.

What Free does not include: Claude Code in the terminal, Cowork, Research access, unlimited projects, or any of the desktop extensions. If you want to use Claude as a developer tool, Free is a starting point only.

Pro: $20 per month

The mainstream individual plan. $17 per month if you pay annually ($200 upfront), or $20 monthly.

Pro adds Claude Code terminal access, Cowork, Research access, unlimited projects, and the desktop extensions. The usage cap is the most important number here, and Anthropic publishes it as approximately 10 to 40 Claude Code prompts per 5-hour rolling window with about 44,000 tokens of headroom shared across chat, projects, and Code.

Who Pro fits: anyone using Claude for a few hours a day across mixed tasks. Solo developers shipping at a normal pace, writers who want a Claude-as-collaborator workflow, anyone evaluating whether deeper tiers are worth it. Most users start here and stay.

Pro hits its limits when you sit in a Claude Code session for more than 90 minutes of dense back-and-forth. If you find yourself rate-limited every day, the upgrade is Max, not the API.

Max: from $100 per month

Max is positioned as the developer power-user tier. Two pricing variants: $100 per month for 5x the Pro usage and $200 per month for 20x. Both include priority routing during peak demand and the same model access as Pro.

The 5x Max tier ($100) is the right choice for developers using Claude Code as a primary tool. It buys back roughly a full working day of friction-free coding before the 5-hour window matters. The 20x tier ($200) makes sense only if you genuinely run multiple parallel Claude Code sessions, or if you are also using Claude heavily for non-coding work in the same account.

Max does not include team features (no SSO, no admin controls, no shared billing). For two or more developers, Team is the better economic choice even if individual usage is lower.

This is where most paid Claudify users sit. If you want to know whether Max delivers on the price, our Claude Code Max plan breakdown goes through real-world rate-limit testing, and Claude Pro vs Claude Max gives you the break-even math for the upgrade specifically.

Team: $20 to $125 per seat

Team plans are billed per seat. Two seat types:

Standard seat at $20 per month (annual) or $25 monthly. Slightly higher usage allocation than Pro and unlocks team admin controls, SSO, audit logs, and the no-training-by-default privacy posture that Anthropic positions as table stakes for any business use.

Premium seat at $100 per month (annual) or $125 monthly. Includes 5x the standard seat allocation and is the closest equivalent to giving each developer a Max-tier individual plan. Premium seats are how teams give engineers serious Claude Code access without the API.

Team minimum is 5 seats. The economics start to work above ~3 developers using Claude regularly, where shared admin and centralised billing pay for themselves. Below that, individual Max subscriptions are cheaper.

Enterprise: $20 per seat plus usage at API rates

The Enterprise tier is structured differently. The base seat is $20 per user per month for the platform (admin tools, SSO, SCIM, audit logs, the compliance API, optional HIPAA-ready setup), but actual model usage is metered at API rates and added on top. There is no flat usage cap.

This works for organisations that want enterprise governance without paying for usage they do not consume. A 200-person company where 30 people use Claude heavily and 170 use it lightly will spend less on Enterprise than on giving everyone a Premium Team seat.

Enterprise also unlocks the spend controls and SCIM provisioning that companies with compliance requirements actually need. SOC 2, HIPAA-ready, and the audit log feed required for SIEM integrations all live here.

If your team is evaluating Enterprise, get in touch with Anthropic sales for pricing. The list price is published, the negotiated rate at any volume is not. For the full break-even math against Teams Premium, our Claude Enterprise pricing guide works through the per-seat-plus-usage model in detail.

API pricing

The API is pay-per-token, billed monthly. Anthropic publishes current rates on their pricing page (model rates change as new generations ship). The general structure has not changed in years:

Sonnet is the default. It is priced for production workloads and most developers building on the API end up here.

Opus costs significantly more per token but unlocks the highest reasoning capability. Use Opus selectively, for tasks where the model quality matters more than the per-call cost.

Haiku is the cheapest and fastest. Designed for high-volume, latency-sensitive workloads where a smaller model is acceptable.

Prompt caching can cut input costs by up to 90 percent for repeated context. Batch processing offers up to 50 percent off in exchange for asynchronous delivery. Both are documented on the API pricing page and are the difference between a sustainable production setup and one that hemorrhages money.

For a Claudify-grade install, the optimisation rule is: use Sonnet by default, switch specific subagents to Haiku for cheap classification work, and never use Opus inside an automated loop. Our Claude models comparison breaks down where each model wins.

Hidden costs nobody talks about

Three things to budget for that do not appear on the pricing page:

Rate-limit recovery time. Hitting the Pro cap costs you 30-60 minutes of waiting. At a developer's hourly rate, that is real money. The Max upgrade pays for itself the first time it prevents a half-hour stall in a critical session.

Subscription drift between plans. Pro and Max are tied to a single email account. If you accidentally start a Claude Code session under the wrong account, you are eating into the wrong quota. Most teams hit this at least once.

API token bloat from poor context discipline. API costs scale with what you put in the input. Sending 50K tokens of unnecessary context per call instead of 2K is the single biggest source of bill shock. This is where memory systems and custom agents earn their cost.

This is exactly where Claudify saves money: the memory architecture, the auditing hooks, and the agent definitions all reduce wasted tokens by structurally preventing the patterns that bloat them.

Which plan should you pick?

A practical decision tree:

You are...	Pick
Trying Claude for the first time	Free
Solo developer, ~1-2 hours/day in Claude Code	Pro
Solo developer, ~3+ hours/day in Claude Code	Max 5x ($100)
Building automation, scripts, or product features	API
Heavy power user, multiple parallel sessions	Max 20x ($200)
Team of 3-10 developers	Team Premium seats
Team of 10+ with compliance needs	Enterprise

The single most common mistake is staying on Pro when you should be on Max. If you spend 30+ minutes a week waiting for the Pro window to roll forward, the $80 upgrade pays for itself in time saved every month.

The second most common mistake is jumping to the API for interactive use. The API has no usage cap but every prompt is itemised on a bill. For a single developer using Claude Code 20 hours a week, the Max subscription is almost always cheaper than the equivalent API spend.

Next steps

Pricing is the floor. The actual return on Claude is determined by how well-configured your install is. A properly-set-up Claude Code project on a Pro plan beats a vanilla install on Max because the model is using its tokens on signal, not noise.

Get Claudify. The operating system for Claude Code that turns any plan into a high-leverage one. Skills, memory, agents, and audit hooks ship out of the box.