Is there an unlimited Claude plan?

Is there an unlimited Claude plan?

No. As of September 2026 there is no unlimited Claude plan. Free, Pro, Max 5x and Max 20x all run on the same two clocks: a session allowance that resets every five hours, and a weekly allowance that resets at a fixed time assigned to your account. Max 20x is the highest cap an individual can buy, and it is still a cap. The only Anthropic product that is metered instead of capped is the API. It charges per token, and it has its own rate tiers and a monthly spend cap per tier.

That is the whole answer. Below are the ways around it, ordered from cheapest to most expensive, followed by the two shortcuts that break Anthropic's terms.

What the limit message means

When you use up the session window in the Claude app you see 5-hour limit reached - resets [time], with the real reset time filled in. Shortly before that you see Approaching 5-hour limit. Anthropic's help center documents both strings. The same window applies in Claude Code, because Claude and Claude Code draw from one shared allowance. On Pro and Max a weekly limit sits on top of the session limit. It applies across all models, and it does not move when the five-hour window resets, so a heavy week can leave you locked out even though the session clock says you are free.

Anthropic does not publish a fixed number of messages per window, so any figure you see quoted is someone's estimate. Usage is measured in tokens, so one long conversation with large attachments empties a window faster than many short questions. Model choice matters too: Anthropic's own Claude Code documentation says Opus costs several times more per turn than Sonnet, and Sonnet more than Haiku. If one afternoon in Claude Code emptied a window that normally lasts all day, how Claude Code spends tokens explains the context growth that did it.

What each plan costs and how much usage it includes

These are Anthropic's list prices in USD as of September 2026. The usage column is Anthropic's own wording: each Max tier is described as a multiple of the Pro plan's per-session allowance.

[
  {
    "plan": "Pro",
    "monthly_price_usd": 20,
    "session_usage_vs_pro": 1
  },
  {
    "plan": "Max 5x",
    "monthly_price_usd": 100,
    "session_usage_vs_pro": 5
  },
  {
    "plan": "Max 20x",
    "monthly_price_usd": 200,
    "session_usage_vs_pro": 20
  }
]

Pro is 20 USD a month billed monthly, or 17 USD a month on an annual plan. Max 5x is 100 USD for 5 times Pro's session allowance, and Max 20x is 200 USD for 20 times. Free is not in the chart because Anthropic only describes it relative to Pro: Pro gives at least five times more usage per session than Free. Every row in that chart has a session limit and a weekly limit. The price buys a taller ceiling, never the absence of one.

Option 1: wait for the window to reset

This costs nothing. The message tells you when the session window opens again, and it is always within five hours of when the window started. The weekly limit resets at a fixed time each week that is assigned to your account. There is no button that resets either one early. Anthropic's help pages give you one free way forward at a limit, waiting, and two paid ways, a bigger plan or usage credits. Resetting early is not on the list, which is why you cannot reset Claude usage limits by asking, and why a new conversation does not help either, since the window is per account rather than per chat.

What you can do while waiting is stop the next window from draining as fast. In Claude Code, /model switches to a cheaper model for routine edits. Start a fresh conversation for a new task, because every turn re-sends the whole context. Do not re-attach a large file in each message. When the 90 percent warning appears, finish the current task rather than starting a new one.

Option 2: turn on usage credits, or buy a bundle

Usage credits are the closest thing to an unlimited plan that Anthropic sells: a subscription that falls through to metered billing when the included usage runs out. They are available on Pro, Max 5x and Max 20x, and they are billed at standard API rates, the same per-token prices the API charges. Turn them on under Settings, then Usage, click Enable, add a payment method, and set a monthly spending cap. The cap can be a fixed amount or set to unlimited. Once credits are on, the limit message changes to 5-hour limit resets [time] - continuing with usage credits and the conversation keeps going. Credits apply to both Claude conversations and Claude Code. There is a daily redemption limit of 2,000 USD, and in most cases the credits do not expire (Japan is an exception, where they expire six months after purchase). The detail of how a credit is metered lives in what usage credits are and how they are billed.

A usage bundle is the prepaid, discounted form of the same thing. You buy a fixed amount in advance and it is drawn down only after your plan's included usage is exhausted. Bundles are available on Pro, Max and Team, and an individual subscriber can buy up to 2,000 USD of them in a month.

[
  {
    "label": "50 USD bundle",
    "face_value_usd": 50,
    "price_usd": 45,
    "saving_pct": 10
  },
  {
    "label": "250 USD bundle",
    "face_value_usd": 250,
    "price_usd": 200,
    "saving_pct": 20
  },
  {
    "label": "1,000 USD bundle",
    "face_value_usd": "1,000",
    "price_usd": 700,
    "saving_pct": 30
  }
]

The smallest bundle costs 45 USD for 50 USD of usage, a 10 percent saving. The largest costs 700 USD for 1,000 USD of usage, a 30 percent saving. One bundle balance is a single pool across the Claude web app, the desktop and mobile apps, Claude Code, Cowork and third-party products. If you already know you will spend the money, a bundle is strictly cheaper than pay-as-you-go credits.

The honest caveat is that credits are not unlimited in price. A Claude Code session on Opus with a large repository can spend real money in an hour, so set the monthly cap before you enable them, and read the API price table below to understand what a token costs.

Option 3: move from Pro to Max 5x, then Max 20x

If you hit the session limit most days, a bigger plan is cheaper than paying for the overflow with credits. Upgrading is prorated: Anthropic charges the difference for the rest of your billing cycle, so you do not have to wait for a renewal date. Max 5x gives five times Pro's session allowance and Max 20x gives twenty times, and both keep the weekly limit. The upgrade raises the ceiling; it does not remove it. Whether the second hundred dollars is worth it depends on the shape of your week, and Max 5x versus Max 20x works through the cases where the cheaper tier is enough.

A rough rule: if you hit the limit once a week, credits are cheaper. If you hit it every day, Max 5x is cheaper. If Max 5x still runs out before lunch, Max 20x plus credits is the most usage any individual plan will give you, and that is the true top of the consumer ladder.

Option 4: point Claude Code at API billing

The API is the one path that is metered rather than capped. You open a Console account, buy prepaid API credits, create a key, and pay per token. Claude Code will use that key instead of your subscription the moment it sees the ANTHROPIC_API_KEY environment variable.

export ANTHROPIC_API_KEY=sk-ant-...
claude

Inside the session, /cost shows the running spend for an API-key session. To go back to the subscription, run /logout and then /login and pick your Claude account. Anthropic's help page warns about this in the other direction: if the variable is set on your machine, Claude Code bills the API even when you thought you were on your plan, and the charge lands on your Console bill rather than counting as plan usage. Unset the variable if you did not mean to use it. Claude Code also offers a softer version of this switch: when you reach a plan limit it presents an option to continue through API credits, and it only does so with your explicit consent.

These are the per-token prices as of September 2026 for the current models. A million tokens is roughly 750,000 English words, and a typical Claude Code turn moves tens of thousands of input tokens because the whole context is re-sent.

[
  {
    "label": "Haiku 4.5",
    "input_usd_per_mtok": 1,
    "output_usd_per_mtok": 5
  },
  {
    "label": "Sonnet 5",
    "input_usd_per_mtok": 2,
    "output_usd_per_mtok": 10
  },
  {
    "label": "Opus 5",
    "input_usd_per_mtok": 5,
    "output_usd_per_mtok": 25
  },
  {
    "label": "Fable 5.1",
    "input_usd_per_mtok": 10,
    "output_usd_per_mtok": 50
  }
]

Sonnet 5 costs 2 USD per million input tokens and 10 USD per million output tokens. Opus 5 is 5 and 25. Cached input is billed at a tenth of the base input price on most models, and less on Fable 5.1, which is why a long Claude Code session costs less per turn than the raw numbers suggest. Picking the model is the biggest lever you have, and which model to use for which job is the short version of that decision.

Metered does not mean uncapped. The API enforces rate limits per model, measured in requests per minute and in input and output tokens per minute. Exceed one and the response is HTTP 429 with error type rate_limit_error and a retry-after header telling you how many seconds to wait. The limits refill continuously, so there is no five-hour wall, but a script that fires requests in a burst still gets 429s. On top of that, each usage tier has a monthly spend cap. When you reach it, requests return the same 429 but with error_code set to enforced_spend_limit_reached and no retry-after header, and usage pauses until 00:00 UTC on the first of next month unless you request a higher limit. New organizations may start in an Evaluation tier with lower limits that rise as the account builds history.

[
  {
    "label": "Start",
    "monthly_spend_cap_usd": 500
  },
  {
    "label": "Build",
    "monthly_spend_cap_usd": "1,000"
  },
  {
    "label": "Scale",
    "monthly_spend_cap_usd": "200,000"
  }
]

Whether the API works out cheaper than a plan depends on how you use it. A steady daily user of Sonnet on Pro is getting a bargain against the token price, while an occasional heavy user pays for a lot of idle days. API billing versus a subscription runs the numbers for both patterns. If you are building something that runs unattended, putting a spend cap and a kill switch on an AI agent matters more than the plan choice, because an agent in a loop is the fastest way to reach any limit on this page.

Option 5: run an open model on your own server

The only truly unlimited option is one where no one meters you. That means a model you run yourself. You cannot do that with Claude; Claude is not available to self-host on any plan, and the weights are not distributed. You can do it with an open-weights model, and the trade is capability for control: a mid-sized open model on a VPS is well below Opus on hard reasoning, and good enough for a lot of routine drafting and summarising.

On a Ubuntu 24.04 VPS with 16 GB of RAM, Ollama gets you a running model in a few minutes.

curl -fsSL https://ollama.com/install.sh | sh
ollama run qwen3:8b

The first command installs Ollama and registers it as a systemd service. The second downloads an 8-billion-parameter model, about 5 GB, and drops you into a chat. Without a GPU it answers at a few tokens per second, which is fine for a chat and slow for an agent. There is no window and no weekly limit. The only bill is the VPS invoice. Running an open model instead of a Claude plan covers which models are worth the RAM and how to wire one into the tools that expect an OpenAI-style endpoint.

What does not work, and breaks the terms

Two shortcuts come up in every forum thread about limits, and both fail.

Sharing one Max login. Anthropic's consumer terms say: "You may not share your Account login information, Anthropic API key, or Account credentials with anyone else. You also may not make your Account available to anyone else." A shared Max 20x login for a small team breaks that clause outright, and it does not even help much, because the whole team drains one five-hour window and one weekly cap together. The product built for that case is the Team plan, where each seat has its own allowance and a Premium seat carries five times the usage of a Standard one; Claude Team plans for a small business explains the seat maths.

Rotating between several personal accounts. Each account is a separate subscription with its own windows, so this is paying twice to reach the same ceiling twice, with your projects and memory split across logins. The terms also forbid accessing the consumer app through scripts or bots, so any tool that automates the hopping is out on its own. A site selling "unlimited Claude" is either reselling API tokens with a markup, in which case buy the tokens yourself, or handing out a shared login, in which case both of you are breaking the clause above.

Which option fits which reader

Someone who hits a limit once in a while should wait, and use the cheaper model in the meantime. Someone on Pro who hits it most days should move to Max 5x. Someone on Max who hits it in bursts should turn on usage credits with a monthly cap, or buy a bundle if the spend is predictable. Someone who wants an itemised bill, or who runs Claude Code unattended, should put it on an API key with a spend limit. Someone who needs no meter at all should run an open model on their own server and accept that it is not Claude.

FAQ

Is Claude Max 20x unlimited?

No. As of September 2026, Max 20x is 200 USD a month for 20 times the Pro plan's per-session allowance, and it has both a five-hour session limit and a weekly limit that applies across all models. It is the highest cap an individual can buy, and adding usage credits on top of it is the most usage a single consumer account can get.

Does the Claude API have usage limits too?

Yes, but they are different in kind. The API bills per token instead of capping messages per window. It enforces per-model rate limits in requests and tokens per minute, returned as HTTP 429 with a retry-after header, and each usage tier has a monthly spend cap that pauses requests until the first of the next month. Only the Custom tier, arranged with sales, has no monthly cap.

Can I buy more usage on the Free plan?

No. Usage credits and usage bundles are available on Pro, Max and Team plans only. On Free, the only options are waiting for the five-hour window to reset or upgrading to Pro, which Anthropic describes as at least five times more usage per session than Free.

After a limit, is it cheaper to buy usage credits or upgrade to Max?

It depends on how often you hit it. Credits are billed at standard API rates, so a rare overflow costs a few dollars and no monthly commitment. If you hit the limit most days, the overflow adds up, and the jump from Pro at 20 USD to Max 5x at 100 USD, which buys five times the session allowance, is cheaper than paying per token for the same amount of work. Upgrades are prorated, so you can switch mid-cycle and test that for one week.