Pick a Claude plan by how you actually use it: how often, for what kind of work, and how often you run into the limit. Once you answer that, the price mostly picks itself.
Imagine a red banner says you have hit your limit, and you upgrade to the most expensive plan that same week. Or you stay on Free for months, hit the limit most afternoons, and decide AI chat is a toy. Both are reactions to a price page, not decisions.
This post compares Free, Pro, and Max the way you would compare phone plans: what you get, where the limits bite, and when paying more makes sense. Prices change often, so check Claude’s pricing page (checked September 2026) before you pay. It is part of Learn Claude from scratch.
The plan ladder in plain English
Here is the ladder for individual Claude plans, the consumer side rather than a company account. Confirm live numbers on the pricing page, because tax, region, and promotions all vary.
- Free: costs $0, as of September 2026 on Claude pricing, and is a real product with real limits, fine for light exploration and occasional drafts.
- Pro: runs about $20 per month if you pay monthly, or about $200 per year if you pay up front, as of September 2026 on Claude pricing (Anthropic’s page has shown that annual path as roughly $17 per month equivalent). It includes more usage than Free, plus the paid feature bundle described below.
- Max 5x: runs about $100 per month, as of September 2026 on Anthropic’s Max plan page. It includes everything in Pro, with roughly 5x Pro’s usage per rolling session window, higher output limits, and priority-style access during busy periods, in the wording the marketing page uses.
- Max 20x: runs about $200 per month, as of September 2026 on Anthropic’s Max plan page, with the same idea at roughly 20x Pro’s usage per session window.
Those multipliers describe relative capacity, not a promise of an exact message count. Long chats, big files, heavier models, code tools, and research-style features drain the pool faster than short questions do, and Anthropic’s own help docs say there is no fixed public message count because tasks cost so differently to run. As of September 2026, that note is on the usage limit page.
How usage limits actually feel
Limits are the part people misread as “Claude got dumber.” Usually the model is fine and the meter is just empty.
The rolling five-hour window
On Claude’s consumer plans, most of your usage resets on a rolling schedule, roughly every five hours. When you hit the cap, you wait for capacity to come back, upgrade, or, on paid plans where it is offered, add usage credits that bill closer to API rates. An API is a way for one program to ask another program for data or an action. There is no simple public “50 messages” card, because a three-line email rewrite costs far less to run than a two-hour coding session. As of September 2026, Anthropic says this on the usage limit page.
Weekly limits on paid plans
Paid plans also use weekly limits so a single user cannot monopolize capacity. You can breeze through a morning sprint and still feel a weekly ceiling if you run heavy tools all week long. Check Settings then Usage in the product when you care about the live bar rather than folklore.
One pool no matter where you open Claude
Web, desktop, mobile, and Claude Code, when it is included on your plan, generally draw from the same plan pool. Running Code in a terminal (the text window where you type commands) does not magically create a second free meter next to your browser chat. That surprise causes a lot of Max upgrades that could have been “stop running two huge sessions at once” instead.
What Pro claims relative to Free
Anthropic’s pricing FAQ has described Pro as providing on the order of at least 5x more usage per five-hour session than Free, with Max multiplying Pro again, either 5x or 20x. As of September 2026, confirm that wording on Claude pricing. Use those ratios as orientation, not a spreadsheet identity, because features and models also change what counts as one unit of work.
If you want the primary source language, read Anthropic’s usage limit article: usage limit best practices.
Free: what you actually get
Free is not a demo video. You can chat on web, iOS, Android, and desktop, depending on current download options, write and edit, generate code, search the web when the feature is enabled, and use a genuinely useful slice of everyday assistant behavior. Memory across conversations, file creation, code execution, connectors, and other extras show up on the marketing comparison table for Free in some periods and not others. The durable truth is simpler: Free is capacity-constrained, and the ceiling shows up the moment you try to make Claude your all-day coworker.
Free fits you if:
- You draft the occasional email, outline, or explanation.
- You are evaluating whether the product style fits your brain before paying anything.
- You already have another paid assistant and want Claude as a second opinion only.
- Your employer has not approved paid consumer AI spend yet, and you stick to non-sensitive practice tasks.
Free will frustrate you if:
- You hit limits mid-project several days a week.
- You need the paid-only tools, Claude Code and similar, as a daily driver.
- You rely on higher model access that Free does not include in the current matrix.
Frustration is data. If Free blocks real work, upgrade for capacity, because that is a genuine need. If Free is fine and you upgrade only because a podcast host called Max “mandatory,” that is fear of missing out (FOMO, the anxious itch to keep up with everyone else), not a real need.
Pro: the default paid seat for most individuals
Pro is where Anthropic expects everyday productivity buyers to land. The headline bundle, as the pricing page currently describes it, includes everything in Free, plus more usage, and a set of named inclusions that matter if you read them as job tools rather than logo stickers.
More usage
This is the feature you will notice first. You get longer sessions before the five-hour window complains, so there is room for iterative drafting: rewrite, then rewrite again, then ask for a skeptical pass. If your pain on Free was “I was in flow and the banner killed it,” Pro is aimed squarely at you.
Claude Code
Claude Code is Anthropic’s coding-oriented product, built around a terminal and related workflows. It is included on paid plans as of the current pricing FAQ. Heavy Code use burns the same plan pool as chat does, so developers who live in a repo all day often outgrow Pro and look at Max, or they split work with API credits instead. Casual “help me fix this query (a question you send to a database)” users may never need Max at all.
Claude Cowork
Cowork is a separate product for agent-style computer work, and its exact scope keeps evolving. Being included on Pro means you can try it without paying a Max tax first. Whether it becomes your daily driver is a question for a later post in this series. For plan choice, ask only one thing: will I open this weekly?
Claude Design and Claude Science
These are specialized products in the Claude lineup, one design-oriented and one science-oriented. Pro marketing lists them as included. If your job is not design systems or scientific workflows, do not buy Pro only for the names. If those are your job, confirm current capabilities on the product pages before you build a process around them.
Research, Projects, and more models
Projects help you keep chats and documents organized for ongoing work, such as a client, a course, or a migration. Unlimited projects on Pro, per current marketing, matters a great deal if you were drowning in one flat chat list. Research-style features help with multi-step information gathering when your plan offers them. Access to more Claude models means you are less stuck on the lightest free default when a hard task needs a heavier engine, with names like Sonnet, Opus, and Fable generations offered depending on your plan.
Claude for Microsoft 365
Pro lists a Microsoft 365 path, which matters at companies standardized on Outlook, Word, and Teams. Mattering is not the same as skipping a security review; connectors and workplace integrations should still follow your employer’s rules. A later post in this series goes deeper on work integrations. Here, the plan takeaway is simply that Pro is the consumer plan that advertises this integration.
Pro is a sensible default if you use Claude several times a week for real deliverables, want Code or Projects, and hit Free’s limits often enough to notice. Pro is optional if you open Claude twice a month for fun outlines and nothing more.
Max 5x and Max 20x: when more capacity is the product
Max is not a different personality. It is mostly more room, plus some priority and early-access framing layered on top of Pro.
- Max 5x (~$100/mo): about five times Pro’s usage per five-hour session window, higher output limits, priority access when traffic is high, and earlier access language for advanced features.
- Max 20x (~$200/mo): the same idea at about twenty times Pro’s session usage. This is power-user territory: long coding days, constant agent-style runs, or roles where waiting on a limit banner costs more than the subscription does.
Buy Max when you can point to a real pattern: “I hit Pro limits on most workdays while doing productive work I already know how to verify.” Do not buy Max because a stranger on the internet insisted 20x is the only serious tier. Plenty of excellent work happens on Pro, and plenty of mediocre work happens on Max, with unverified SQL and confident wrong numbers to show for it. SQL is the language for asking a database questions.
If you are a developer comparing Max to pure API spend, run the math for your own token volume. A token is a small chunk of text, often part of a word, and API use is billed by the chunk. API pricing is public on the platform docs; consumer Max is a flat-ish subscription with shared limits. Neither one is universally cheaper than the other.
A comparison table you can actually use
Treat these values as orientation, checked at one point in time. Confirm the live numbers on claude.com/pricing.
| Plan | Ballpark price | Usage orientation | Best fit |
|---|---|---|---|
| Free | $0 | Light daily; easy to hit walls | Trying Claude, light drafts |
| Pro | ~$20/mo or ~$200/yr | Much more than Free; still finite | Regular work, Code, Projects |
| Max 5x | ~$100/mo | ~5x Pro per ~5h window | Daily heavy use, frequent Pro caps |
| Max 20x | ~$200/mo | ~20x Pro per ~5h window | All-day power workflows |
The API Console is a different product wallet
Two Anthropic-related logins confuse people for months at a time.
- claude.ai subscription (Free / Pro / Max): human chat and the included products, metered by plan limits and optional usage credits on paid plans.
- Claude Platform / Console (API): keys for applications, automation, and builders. You pay per token, plus related tool fees, for models such as Haiku, Sonnet, Opus, and Fable generations. Pricing tables live on the platform docs and the API section of the pricing page.
You can hold both accounts at once, and many developers do: Pro or Max for interactive Code and chat, Console keys for apps. Paying for Pro does not automatically fund unlimited API traffic. An API budget does not automatically add every consumer chat perk to your account either.
If your job is “I talk to Claude in a window,” you can ignore Console until a later post covers it. If your job is “I ship a feature that calls a model,” start with the platform docs and finance’s cloud process, not only a personal Max card.
Team and Enterprise: when the org should buy
Personal plans are built for individuals. As soon as several people need shared admin, single sign-on (SSO, one login that covers every company tool), central billing, or clearer training defaults, look at organization plans instead.
Anthropic’s pricing page currently describes two organization tiers:
- Team: for groups, with marketing citing bands such as 2 to 150 people. Standard and Premium seats carry different usage. Central billing, SSO, admin controls for connectors, Claude Code and Cowork inclusions, and default “no model training on your content” style enterprise messaging round it out. Standard seat pricing has been shown around $25 per seat per month billed monthly, or $20 per seat per month on the annual-equivalent path; Premium runs much higher, on the order of $100 to $125 per seat per month depending on annual versus monthly billing. Verify the live numbers before you commit.
- Enterprise: larger scale, with seat price plus usage at API rates in the self-serve framing, and deeper admin, automatic account provisioning (SCIM, a standard way to add and remove user accounts across systems), audit logs, retention controls, network controls, and compliance-oriented options, including healthcare-privacy compliance (HIPAA, the US law covering medical privacy) where it applies. Sales-assisted paths exist for complex contracts.
Smell test: if five employees each buy Max on their personal cards and paste customer data into personal accounts, that is not a productivity win. That is a shadow IT pile. Team or Enterprise plus policy training is the adult move, even when the sticker price looks higher than five Pro seats added together.
Plan chooser: pick without FOMO
Walk these questions in order and stop at the first plan that fits.
- Is this for work with company data? If yes, ask IT or your manager which Claude path is approved. Do not invent a personal Max workaround for regulated data.
- How many days per week do you hit Free’s limits while doing useful work? If rarely, stay on Free. If most weeks, consider Pro.
- Do you need Claude Code, Projects organization, Research, or Microsoft 365 integration as a regular tool? That points to Pro, or an org plan, rather than Free heroics.
- On Pro, do you still hit limits on most full workdays? Track it for two weeks. If yes, Max 5x is the usual next step. If you are a full-day agentic (working in many steps on its own) or coding user and 5x still bricks you, Max 20x enters the chat.
- Are you building software that calls models in production? Budget Console or API spend separately, since a chat plan is not built to replace engineering cost controls.
- Are multiple people involved? Price Team or Enterprise against the chaos of personal cards, and include admin time in that comparison.
Notice what is missing from that list: “which model is winning benchmarks this week?” Benchmarks are interesting, but they rarely justify a $200 personal subscription if your actual job is rewriting status updates twice a week.
Cost scenarios (toy math, not invoices)
These are teaching scenarios with round numbers. Your taxes, currency, and promotional pricing will differ from what is shown here.
Scenario A: curious analyst
Three sessions a week, short summaries, no Code. Free is enough for this. If a quarterly deep-dive week feels cramped, a single month of Pro can work as a temporary boost, and canceling paths exist in billing settings; always follow the live cancel instructions there.
Scenario B: an individual contributor who writes and codes lightly
Daily chat, weekly Code, Projects for two active workstreams. Pro at roughly $20 per month is the default here. Annual billing at roughly $200 makes sense only if you already know you will keep it a full year and the discount still exists when you subscribe.
Scenario C: full-time coding with agents
Claude Code open most of the day, large repos, and frequent use of higher-tier models. Max 5x or 20x may be the rational call here. Compare that against API-heavy workflows and any company-provided seats. Keep verifying code either way; a higher plan does not reduce the need to check AI-written SQL or review diffs.
Scenario D: ten-person team, mixed roles
Ten personal Pro seats already run $200 per month before anyone buys Max, and that number ignores the added shadow-data risk. Team standard seats may land near a similar monthly number, with SSO and admin thrown in, depending on discounts and the mix of seat types you choose. Run the real quote before deciding, since the point here is organizational design, not just arithmetic.
Common plan mistakes
- Upgrading before learning prompts. A vague one-liner wastes Max as efficiently as it wastes Free, so practice iteration on Free for a week with non-sensitive tasks first.
- Ignoring the shared pool. Chat plus Code plus mobile “quick questions” is one meter, so stagger heavy jobs if you are close to a limit.
- Using personal Pro for secret work data. Plan tier does not equal compliance. Policy does.
- Assuming annual is always better. Annual makes sense once you are sure. If you are still experimenting, monthly billing reduces regret.
- Mixing API bills with chat bills in your head. Keep separate budgets and separate owners where possible.
- Chasing model names as a substitute for process. Heavier models help hard tasks, but they also burn through limits faster, so match the engine to the job.
- Treating the assistant as a source of record because you pay more. Payment buys capacity and features, not truth. Liability stays with a human either way.
Practice: a two-week plan experiment
You do not need a spreadsheet temple for this. You need honest notes.
- Stay on your current plan for ten workdays. Each day you use Claude, jot down minutes used, the main task type (write, summarize, code, research), and whether you hit a limit banner.
- Mark sensitive versus non-sensitive tasks. If sensitive work appeared on a personal account, flag it for policy cleanup regardless of plan.
- Count limit events. Zero to one mild limit in two weeks means stay put, or move to a light Pro plan. Several mid-task interrupts while doing real work point to Pro or Max, depending on severity.
- List the features you actually opened. If you never opened Code, Cowork, Design, or Science, do not let those logos alone drive a Max purchase.
- Decide in one sentence. For example: “Pro monthly for three months, because Free blocked two client-facing drafts per week.” Revisit the decision after the quarter ends.
Before any upgrade, it is worth revisiting the boundaries from the earlier post in this series: paying more never makes invented metrics safe for a board deck.
Age, terms, and other fine print worth knowing
Consumer Claude is for users 18 and older under current Anthropic consumer terms; confirm the live legal page for the exact wording. Refunds are limited and depend on your region and storefront, whether that is the web, Apple, or Google. Downgrades usually apply at the end of the current period, while upgrades often apply immediately with proration-style credits. None of that replaces reading the billing screen in front of you when you click.
Feature matrices on marketing pages shuffle over time. A capability that is free today might move behind a paywall tomorrow. A model name in a blog post might differ from the picker inside your own account. When in doubt, the account you are logged into is the source of truth for what you personally can click, and the legal and pricing pages are the source of truth for what Anthropic claims to sell.
Quick recap
- Free is real but capacity-tight. Pro (~$20/mo or ~$200/yr) is the usual individual paid seat. Max 5x (~$100) and Max 20x (~$200) buy more usage on top of Pro.
- Usage resets on a rolling window about every five hours; paid plans also carry weekly limits, and every surface shares one pool.
- Pro’s headline extras include more usage, Claude Code, Cowork, Design, Science, Research, Projects, more models, and a Microsoft 365 path; confirm the live list before you buy.
- API Console billing is separate from chat subscriptions.
- Teams should prefer org plans and written policy over a pile of personal Max cards for work data.
- Choose with two weeks of limit notes, not with FOMO. Re-check claude.com/pricing before you pay.
- Paying more never makes Claude a source of record. You still own the numbers.
Series notes
This is Part 2 of Learn Claude from scratch (plan pricing and usage limits). The earlier post placed company, model, and product in separate boxes. The next post walks first sessions on web and mobile, and later posts cover Projects, everyday jobs like writing, summarizing, and planning, privacy habits, work integrations, and when Claude is the wrong tool for the job. Specialized product deep dives for Code, Cowork, Design, and Science get their own tracks, so this post can stay focused on money and capacity. If you write SQL with any assistant, keep the verification loop from how to check AI-written SQL; plan tier never replaces that loop.
Sources
- Claude plans and pricing (Free, Pro, Max, Team, Enterprise, API overview): https://claude.com/pricing
- What is the Pro plan: https://support.claude.com/en/articles/8325606-what-is-the-pro-plan
- What is the Max plan: https://support.claude.com/en/articles/11049741-what-is-the-max-plan
- Usage limit best practices (five-hour sessions, weekly limits, shared pool): https://support.claude.com/en/articles/9797557-usage-limit-best-practices
- Manage usage credits for paid Claude plans: https://support.claude.com/en/articles/12429409-manage-usage-credits-for-paid-claude-plans
- Claude Platform API pricing docs: https://platform.claude.com/docs/en/about-claude/pricing
- Claude Code product page: https://claude.com/product/claude-code
- Claude Cowork product page: https://claude.com/product/cowork
- Consumer terms of service: https://www.anthropic.com/legal/consumer-terms
- Analytics Made Simple: What are LLMs, ChatGPT, generative AI, and more: https://analyticsmadesimple.com/analytics/what-are-llms-chatgpt-generative-ai-and-more/
- Analytics Made Simple Practical AI series: https://analyticsmadesimple.com/series/practical-ai/
- Analytics Made Simple: How to check AI-written SQL: https://analyticsmadesimple.com/how-to-check-ai-written-sql/
Keep going
Same lessons in your feed
Short diagrams, hooks, and weekly tutorials on Substack, Instagram, X, and Facebook.
