Is Claude free? The free tier explained (2026)
Read on October 26, 2026. Every plan, price and usage limit below was read that day from Anthropic's pricing page, its plan and usage help articles, and its model documentation (sources and method at the end). This is a read of what Anthropic publishes, not a hands-on test of a free account.
Yes, Claude is free. Anthropic runs a standing free plan at $0, and in 2026 that plan is genuinely capable: you get Claude Sonnet and Haiku, web search, memory across conversations, file creation, code execution, connectors, Artifacts, and Claude Design, Slides and Docs. The catch is volume and the flagship models: free usage runs on a limited rolling five-hour allowance, and the most capable Opus and Fable models plus Claude Code and Research are reserved for paid plans. This guide decodes exactly what the free tier includes, what it gates, and the precise point where the $20/mo jump to Pro starts to pay for itself. Every figure here was read from claude.com/pricing and Anthropic's help and model pages on October 26, 2026.
Price to start
"Free for everyone" on Anthropic's pricing page.
email sign-upCapabilities included
Plus Projects capped at 5. Twelve more are gated to Pro or Max.
57% coverageUsage per 5-hour session
Pro is at least 5x Free. Max is 5x or 20x Pro. No fixed message count.
resets every 5 hoursIs Claude really free, or is it a trial?
Claude is really free, and it is a standing plan rather than a trial. When you sign up at claude.ai you land on the free tier with no countdown and no trial clock. Anthropic's pricing page shows Free as the first card on its Individual tab, above Pro and Max, with Team and Enterprise on their own tab. Its help article says anyone in a supported location can use the free version by signing up with an email address, and its plan table marks credit card payment "n/a" for Free. If you later cancel Pro or Max, the pricing page says your account moves to the Free plan once the period ends, and your chats, projects and files stay with your account.
That matters because "free AI" usually means one of two lesser things: a time-boxed trial that converts to a paid plan, or a stripped demo that cannot do real work. The Claude free tier is neither. Sonnet and Haiku are listed on every plan, and the free card carries features many people still assume are paid: web search, memory, file creation, connectors and Claude Design. The constraint is not capability. It is how much of that capability you can use before the rolling allowance pauses you.
The constraint is not capability. It is how much of that capability you can use before the rolling allowance pauses you.Really free?
The Claude free tier decoded: a capability matrix
The fastest way to understand the free plan is to map it against the whole paid Claude surface, capability by capability. The matrix below is the centerpiece of this guide. It takes every capability Anthropic tabulates for its individual plans, 29 of them, and shows which gate opens each one. That gives a single computed number, a coverage score, that answers "how much of Claude do you actually get for nothing?"
Which gate opens which capability
29 capabilities from Anthropic's plan tables, grouped by the cheapest plan that carries them. Read 2026-10-26.
The same matrix as a text table (29 rows)
| Capability | Free | Pro | Max |
|---|---|---|---|
| Chat on web, mobile, desktop | Yes | Yes | Yes |
| Claude Sonnet | Yes | Yes | Yes |
| Claude Haiku | Yes | Yes | Yes |
| Web search | Yes | Yes | Yes |
| Memory across conversations | Yes | Yes | Yes |
| Files and code execution | Yes | Yes | Yes |
| Artifacts | Yes | Yes | Yes |
| Connectors | Yes | Yes | Yes |
| Skills | Yes | Yes | Yes |
| Claude Design | Yes | Yes | Yes |
| Claude Slides | Yes | Yes | Yes |
| Claude Docs | Yes | Yes | Yes |
| Voice mode | Yes | Yes | Yes |
| Desktop extensions | Yes | Yes | Yes |
| Incognito chats | Yes | Yes | Yes |
| User preferences | Yes | Yes | Yes |
| Projects | Up to 5 | Yes | Yes |
| Claude Opus | No | Yes | Yes |
| Research | No | Yes | Yes |
| Claude Code | No | Yes | Yes |
| Claude in Chrome | No | Yes | Yes |
| Claude for Microsoft 365 | No | Yes | Yes |
| Claude for Microsoft Outlook | No | Yes | Yes |
| Claude Dashboards (beta) | No | Yes | Yes |
| Claude Science | No | Yes | Yes |
| Usage credits | No | Yes | Yes |
| Claude Fable | No | Usage credits | 50% of weekly limits |
| Priority access at high traffic times | No | No | Yes |
| Monthly API credits | No | No | $100 (Max 5x) or $200 (Max 20x) |
The arithmetic is the point. Of 29 capabilities in Anthropic's plan tablesread 2026-10-26, the free plan fully includes 16, caps 1 (Projects, up to 5) and withholds 12. Counting the capped row as half, that is 16.5 of 29, a 57 percent coverage score. Pro scores 91 percent (26 full, Fable on usage credits counted as half, 2 withheld) and Max 98 percent. Everything an ordinary person needs from a chatbot, reasoning, writing, search, file work, slides and design, sits inside the free 57 percent. The paid 43 percent is volume, the flagship models and the work-tool surfaces: Claude Code, Research, Claude in Chrome and Microsoft 365.
Coverage by plan, to scale
Each bar is the same 29 capabilities. Width = count of capabilities.
One correction to an earlier edition of this page, which scored the free plan at 75 percent on a hand-picked 12 rows: counting every row Anthropic tabulates is a stricter test, and two calls changed. Claude Design, Slides and Docs are included on the Free plan (the earlier edition listed Design as excluded), and connectors are marked Yes on Free with no limit flag.
Which model does the Claude free plan give you?
The free tier gives you Claude Sonnet and Claude Haiku. Opus is marked No on Free and Yes on Pro and Max, and Fable, above Opus, is paid-only: Pro reaches it through usage credits, Max within 50 percent of its weekly limits. Anthropic's support article on Fable says it plainly: "Fable 5 isn't available on the Free plan." So the single most important thing a free user does not get is the most capable models.
One detail changed since the June edition. That edition named Haiku 4.5. Anthropic's pricing page now lists Haiku 5.5 as the latest Haiku and Haiku 4.5 among its legacy models, and the Free column of its plan table says only "Sonnet" and "Haiku" without a version. We therefore do not say which version the free plan serves today. The current lineup on Anthropic's pricing page is Fable 5.1, Opus 5.5, Sonnet 5.5 and Haiku 5.5.
In practice the gap is smaller than it sounds for everyday use. Anthropic's model documentation describes Sonnet 5.5 as "the best combination of speed and intelligence," and it handles coding, analysis, long-form writing and research at a level that was flagship-tier not long ago. You feel the absence of Opus on the hardest problems: deep multi-step reasoning, large agentic tasks, the cases where you want maximum intelligence regardless of speed. For those, Pro unlocks the line described in our Claude Pro vs Free breakdown.
| Model | Free | Pro | Max | Anthropic's one-line description |
|---|---|---|---|---|
| Claude Haiku (latest: 5.5) | Yes | Yes | Yes | "For high-volume, latency-sensitive tasks" |
| Claude Sonnet (latest: 5.5) | Yes | Yes | Yes | "The best combination of speed and intelligence" |
| Claude Opus (latest: 5.5) | No | Yes | Yes | "For long-running agentic coding and knowledge work" |
| Claude Fable (latest: 5.1) | No | Usage credits | 50% of weekly limits | "For demanding reasoning and long-horizon agentic work" |
What are the Claude free usage limits?
Here is the honest answer most listicles get wrong: Anthropic does not publish a fixed free message count. Its pricing page says the amount you can do "depends on the length and complexity of your conversations, the model you choose, and the features you use, so there's no fixed message count." Any hard figure circulating online, such as "15 messages a day," is a guess rather than a primary-sourced limit. What Anthropic does publish is the shape of the limit: while you use the free plan, there is a session-based usage limit that resets every five hours, and the number of messages "will vary based on demand." What eats the allowance, per Anthropic's usage-limit guidance:
- Message and file length, since longer prompts, replies and attachments consume more per turn
- Conversation length, because the whole thread is carried forward, so a long chat quietly burns more usage
- Tools, such as web search and research, which cost more than plain text
- Model and effort level, where heavier work draws down faster than light queries
- Artifacts and multi-step tasks, such as running code, creating files or browsing websites
- Current demand, since free capacity varies with load
The usage gate, drawn to scale
Usage per 5-hour session, in Free-sized allowances. Every bar is a floor ("at least").
The one number Anthropic does stand behind is relative, not absolute. Its pricing page says Pro gives you at least 5x more usage per 5-hour session than Free. That makes Free the one-times baseline, and the figure above draws the whole ladder to scale. Free usage returns on a rolling five-hour session window instead of a midnight reset. Anthropic adds that paid plans carry weekly limits on top, and that it "may limit your usage in other ways, such as weekly and monthly caps or model and feature usage, at our discretion," so treat the five-hour window as the main limit, not the only one. Our companion guide goes deeper on the mechanics in Claude free plan limits, decoded.
What do you not get on the Claude free plan?
The free plan's gates are about volume, the flagship models and the work-tool surfaces, not core chat capability. Mapping the free tier against Pro and Max on Anthropic's pricing page, here is precisely what stays behind the paywall:
- The Opus models, available from Pro upward, and Fable, which Pro reaches through usage credits and Max within 50 percent of its weekly limits
- Claude Code, the agentic coding tool included in all paid plans and excluded from Free. Anthropic's setup docs say it plainly: Claude Code requires a Pro, Max, Team, Enterprise, or Console account. A developer who wants it without a chat subscription uses a Console account billed per token (see the API section)
- Research, the multi-step research mode, plus Claude in Chrome, Claude for Microsoft 365 and Outlook, Claude Dashboards (beta) and Claude Science, all marked No on Free
- More than five Projects. Free is "Up to 5"; the Pro column says Yes with no cap listed, so "unlimited" is our shorthand, not Anthropic's word
- At least 5x the usage per session, the single biggest practical difference for anyone who uses Claude daily
- Usage credits, the pay-as-you-go top-up that lets a paid plan keep working past its limit
- Priority access at high traffic times and monthly API credits ($100 on Max 5x, $200 on Max 20x), both Max only on the plan table. Anthropic's Pro help article still lists "Priority access to Claude during high-traffic periods" as a Pro benefit, so the two pages disagree; do not upgrade to Pro for it alone
Notice what is not on that list. Web search, memory, file creation, code execution, connectors, Artifacts, Claude Design, Slides and Docs, and Sonnet itself are all free. The modern paywall is narrow and specific: the flagship models, the developer and office tools, and the volume.
The modern paywall is narrow and specific: the flagship models, the work tools, and the volume.Not free
Free vs Pro: when is it actually worth upgrading?
The upgrade decision is simple once you stop comparing feature lists and start comparing how often you hit the wall. Pro costs $20/mo billed monthly, or $17/mo on annual billing ($200 up front)read 2026-10-26, and it adds at least five times the usage, the Opus models, Claude Code, Research, Claude in Chrome, Microsoft 365 and Outlook, Dashboards and Science. The question is whether you need any of that.
Free ($0)
Standing plan, email sign-up
The 1x usage baseline
Pro ($20/mo)
$17/mo on annual, $200 up front
At least 5x the usage
| Factor | Free | Pro ($20/mo) | Max (from $100/mo) |
|---|---|---|---|
| Price | $0 | $20/mo ($17 annual) | From $100/mo, billed monthly |
| Models | Sonnet, Haiku | + Opus (Fable on usage credits) | + Fable within 50% of weekly limits |
| Usage per 5-hour session | 1x baseline | At least 5x Free | 5x or 20x Pro |
| Weekly limits | Not listed for Free | Yes, on top | Yes, on top |
| Claude Code | No | Included | Included |
| Projects | Up to 5 | Yes | Yes |
| Research | No | Yes | Yes |
| Claude in Chrome, Microsoft 365 | No | Yes | Yes |
| Context window | Up to 1M, varies by model | Up to 1M, varies by model | Up to 1M, varies by model |
| Priority at peak demand | No | Unclear: plan table says No; Pro help article says yes | Yes |
| Monthly API credits | No | No | $100 (5x) or $200 (20x) |
Stay free if you use Claude a few times a day for chat, search, writing, slides and the occasional code or file task. The plan is built for exactly that and you will rarely touch the gates. Upgrade to Pro the moment you are hitting the usage wall more than a couple of times a week, or you genuinely need the Opus models, Claude Code, Research or the Chrome and Microsoft 365 tools. The full ladder, Pro, Max, Team and Enterprise, is laid out in our every Claude plan compared guide, and if you are weighing the next rung up, see what Claude Pro actually gets you in 2026 and whether Claude Max is worth it.
Decision tree: free or Pro for you
Answer in order and stop at the first yes. The rules come straight from the matrix: a gate that is a dash in the Free column is a reason to pay, and a gate you never touch is not.
Do you need Claude Code, Research, Claude in Chrome, Claude for Microsoft 365 or Opus?
YESTake Pro. Every one of those is a dash on Free and a Yes on Pro. $20/mo or $17 annualNOGo to question 2.Do you hit the free usage limit more than a couple of times a week?
YESTake Pro. At least 5x the usage per session is the one gate that scales. 5x+ per sessionNOGo to question 3.Do you need more than five Projects?
YESTake Pro. Free is capped at five. up to 5 on FreeNOGo to the end.Stay on Free.
Three no answers mean you sit inside the 16 free capabilities and the 1x allowance fits how you work. Already paying for Pro and still hitting the limit all day? The next rung is Max, from $100/mo, which is 5x or 20x Pro plus priority access.
A verdict for each kind of reader
A student or casual user
FreeSonnet, web search, memory, files and Claude Docs cover everyday work, and you rarely reach the session limit.
A writer or analyst using it daily
Free, then Pro at the wallThe only gate that bites is usage. Pro is at least 5x Free for $20, at most $4.00 per Free-sized allowance.
A developer who wants Claude Code
Pro or a Console accountThe free plan does not include Claude Code; Anthropic's docs list Pro, Max, Team, Enterprise or Console as the accounts that work.
A researcher or Microsoft 365 user
ProResearch, Claude in Chrome and Claude for Microsoft 365 and Outlook are No on Free and Yes on Pro.
Someone who lives in Claude all day
Pro, then Max from $100/moMax is 5x or 20x Pro with priority access at high traffic times; move up only when Pro runs out.
A team
Team or EnterpriseThose sit on their own pricing tab; see our Team and Enterprise pricing guide.
Is there a free Claude API or affiliate deal?
Two questions get bundled in here, and both have clean answers. First, the Claude API is pay-per-token and separate from the free chat plan. Anthropic's model documentation prices Sonnet 5.5 at $2 per million input tokens and $10 per million output tokens, and Haiku 5.5 from $0.10 input and $0.50 output (read October 26, 2026). We found no free API allowance in those docs and did not verify whether new developer accounts receive starting credit, so we do not claim one. If you want to build on Claude or run Claude Code without a chat subscription, you pay per token at the rates in the Anthropic API pricing docs. That is not "free," but it can be cheap for light use.
Second, and this is the one that saves people from scams: Anthropic publishes no coupon, promo code or standing discount for individual Claude plans, and we found no affiliate or referral program for them. Its Pro help article says: "We occasionally run limited-time promotions, but we don't have any standing discounts available upon request," and that its Support team "cannot issue one-off discounts or coupons." The only built-in saving is annual billing, which Anthropic's pricing page says is "designed to cost less than monthly subscriptions." Anything advertising a "free Claude Pro code" or a referral kickback is not from Anthropic. The only real way to use Claude for nothing is the official free plan this guide describes.
How to get the most out of the Claude free plan
If you want to live on the free tier and rarely hit the wall, the levers are usage discipline, not tricks. They follow Anthropic's own usage-limit guidance:
- Match the model to the task. Use Sonnet for reasoning, coding and writing, and switch to Haiku for quick factual lookups. Lighter work draws down the allowance more slowly.
- Start a fresh chat per task. Long conversations are carried forward on every turn and count against your session usage, so a clean thread for each new job stretches the free allowance further than one sprawling mega-chat.
- Group and scope your questions. Anthropic's tips are to gather your questions before you start, group related ones into one message, make requests specific, and refer back to earlier parts of the conversation instead of repeating yourself. A tight question costs far less than a dump-everything prompt.
- Put documents you reuse into a Project. Anthropic notes that cached content counts less against your limits, and Free includes up to five Projects.
- Wait out the reset. If you do hit the wall, the session window resets on its own within five hours. There is no action to take and no penalty; you just come back later.
Do those things and the free plan covers most personal and light-professional use indefinitely. The day they stop being enough, that is your honest signal to upgrade, and not a moment before.
Get the Claude Pricing and Plans Cheat Sheet (PDF)
Every Claude plan and the current API rates on one page, plus the gotchas that change what you actually pay. The PDF's figures were last verified September 27, 2026; this page was re-read October 26, 2026.
The bottom line: is Claude free?
Yes, Claude is free, and properly so. The free plan costs $0, you sign up with an email address, and in 2026 it is a real tool rather than a demo: Claude Sonnet and Haiku, web search, memory, file creation and code execution, connectors, Artifacts, and Claude Design, Slides and Docs, all under a usage limit that resets every five hours. Measured against the 29 capabilities in Anthropic's own plan tables, the free tier includes 16, caps 1 and gates 12, a 57 percent coverage score. The gated 43 percent is specific: the Opus and Fable models, Claude Code, Research, Claude in Chrome and the Microsoft 365 tools, and at least 5x the per-session usage, all of which start at Pro for $20/mo ($17 annual). Stay free if you use Claude casually; upgrade only when you are hitting the wall repeatedly or you genuinely need a flagship model or a work tool. And ignore any "free Claude Pro code," because Anthropic publishes no coupon or standing discount for individual plans and the only real free Claude is the plan above.
How this page was read, and how the finding is computed
What was done. On October 26, 2026 every plan, price and limit on this page was read from Anthropic's own pages. We did not run a free account for this revision, so nothing here is a hands-on test, and the page carries no "tested" date for that reason. Sources, all read that day:
- claude.com/pricing: plan cards, the feature and model tables, the usage-limit and Claude Code answers, current model names.
- What is the Pro plan?: Pro price, priority access wording, the promotions and discounts statement.
- Usage limit best practices and Get started with Claude: what consumes usage, the five-hour session reset on the free plan, email sign-up.
- Claude Fable models on your plan: Fable on Free, Pro and Max.
- Models overview: current model lineup, one-line descriptions, per-token prices.
- Claude Code setup: which accounts can authenticate Claude Code.
Which rows are counted. Every capability that Anthropic's feature and model tables list for the individual plans, one row each as Anthropic lists them (Design, Slides and Docs are three rows; Microsoft 365 and Outlook are two). Left out: rows that are No on every individual plan (@Claude, project sharing, enterprise search, Claude Motion beta), administration and payment rows, and the context window, which is "Up to 1M, varies by model" on every plan. That leaves 29.
Scoring. A capability that is simply included scores 1. A capped or metered one scores half: Projects on Free (up to 5), and Fable on Pro (usage credits) and Max (50 percent of weekly limits). A withheld one scores 0.
- Free: 16 full + 1 capped + 12 withheld = 16.5 / 29 = 56.9 percent, shown as 57.
- Pro: 26 full + 1 metered + 2 withheld = 26.5 / 29 = 91.4 percent, shown as 91.
- Max: 28 full + 1 metered = 28.5 / 29 = 98.3 percent, shown as 98.
- The 12 withheld on Free: Opus, Research, Claude Code, Claude in Chrome, Claude for Microsoft 365, Claude for Microsoft Outlook, Claude Dashboards (beta), Claude Science, usage credits, Fable (10 opened by Pro) and priority access and monthly API credits (2, Max only).
- Usage price: Pro is "at least 5x" Free for $20/mo, so $20 / 5 = $4.00 at most per Free-sized allowance; $17 / 5 = $3.40 on annual billing. Max multiplies Pro, so its multiples (25x and 100x of Free) are floors too. The Max 20x price is not listed on the page, so no price per allowance is computed for it.
What we could not confirm. Which Sonnet and Haiku version the free plan serves (the table names none). Whether extended thinking is a free feature: Anthropic's "Get started" article says you can choose whether Claude uses it, without tying that to a plan, and the pricing tables do not list it, so we no longer count it. Whether free accounts have weekly limits (the pricing page says paid plans add them). Whether new API accounts receive starting credit. Reuse: cite the matrix with a link to this page.
Standards and research behind this page
Claims about what an AI tool can and cannot do are easiest to check against the standards bodies and research groups that measure these systems rather than against any vendor's own description. The sources below are the primary ones, linked so a reader can verify a claim here without taking this page's word for it.
- NIST AI Risk Management Framework is the US federal reference for evaluating and governing AI systems, including how their capabilities and limits should be characterised.
- NIST AI RMF Knowledge Base sets out the measurable characteristics of a trustworthy AI system, which is the vocabulary these comparisons use.
- Stanford HAI AI Index publishes the annual measured benchmarks on model capability, cost and adoption that vendor claims are checked against.
- National Science Foundation AI programme funds and documents the underlying research these products are built on.
- arXiv cs.AI carries the primary papers behind most model capability claims, usually months before a product mentions them.
- FTC Endorsement Guides governs how a review or recommendation must disclose a material connection, which is why the disclosure on this page exists.
- US Copyright Office states what copyright covers, which decides who owns the output of a generative tool.
- Section 508 defines the federal accessibility requirements a tool must meet to be usable in public-sector work.
This page compares tools and does not endorse one. Capabilities and pricing change frequently, and a reader should confirm current behaviour with the vendor before relying on it.