Claude Max $100 vs Ultra $200 per Month

Is there a cheaper Claude Max plan?

Max costs $100 a month, half the price of Ultra at $200 a month, as of October 2026. The former Max 20x plan is now called Ultra, and existing subscribers move to it automatically with no price change. Both plans give you the same models. The higher price buys more weekly usage and no five-hour limit for verified accounts.

[
  {
    "plan": "Pro",
    "price_usd": 20,
    "usage_vs_pro": 1,
    "usd_per_pro_unit": 20
  },
  {
    "plan": "Max",
    "price_usd": 100,
    "usage_vs_pro": 5,
    "usd_per_pro_unit": 20
  },
  {
    "plan": "Ultra",
    "price_usd": 200,
    "usage_vs_pro": 10,
    "usd_per_pro_unit": 20
  }
]

Anthropic publishes the multipliers against Pro: Max gives 5 times Pro's usage every five hours and every week, and Ultra gives 10 times Pro's usage every week. Divide the price by the multiplier and the shape of the offer is clear. Max costs $20 for each Pro-sized weekly allowance, which is exactly what Pro itself costs. Ultra costs $20 for the same unit. For subscription usage alone, both plans have the same price per Pro-sized weekly allowance; Ultra doubles the weekly allowance and removes the five-hour limit for verified accounts.

Max and Ultra also now include monthly Claude API credits: $100 on Max and $200 on Ultra. New subscribers can claim them after seven days by linking a Claude Console organization from Settings > Billing. Unused credits expire at the end of each billing cycle. They cover building apps and agents on the Claude Platform, not interactive Claude Code or usage credits after you hit your subscription limits.

Both plans include Claude Code, the coding agent that runs in your terminal, and Cowork, which does the same kind of multi-step work on ordinary files. As of October 2026 Anthropic is rolling Cowork into the main Claude app on Pro, Max and Ultra, under the line "Claude Cowork is now just Claude". Both plans list early access to new features. Pro includes Claude Code and Cowork as well, so access to those products is the same across all three plans. If you assumed the terminal agent was a Max exclusive, Claude Code is included with Pro too, and that piece also covers what shares your usage window and when an API key quietly bills you instead. For the whole ladder from Free to Enterprise, read which Claude plan you actually need. If the $200 question in front of you is which vendor to give it to rather than whether to upgrade from Max to Ultra, Ultra against ChatGPT Pro at the same price sets the two $200 plans side by side. ChatGPT's $200 plan is no longer OpenAI's top individual tier, though: as of October 2026 OpenAI also sells Pro 500 at $500 a month, which it says offers the highest included usage of its three Pro plans.

What does the Claude Max usage multiplier measure?

Max's multiplier applies to both five-hour sessions and weekly usage, against Pro. A session starts with your first message and resets in full five hours later. Ultra's multiplier measures weekly usage, and verified Ultra accounts have no five-hour limit. The chart compares weekly allowances, so it is worth knowing what Pro's allowance actually buys you before you decide whether you need five times or ten times as much.

Sitting on top of Max's session limit is a weekly ceiling. Pro carries one weekly limit that applies across all models. As of October 2026, Anthropic's help pages describe two weekly limits for Max and Ultra: a general one that applies across all models, and a separate one for Fable models. In practice Max and Ultra can spend up to 50% of their weekly usage on Fable at no extra cost, and Fable draws the shared weekly limit down faster than other models, so the Fable cap is a share of your weekly allowance and not extra on top of it. The weekly window resets at a fixed day and time assigned to your account, and that reset time does not move when you start or stop using Claude. Settings > Usage on claude.ai shows your usage and the next reset time.

Anthropic publishes multipliers and not token counts, because a session holds no fixed amount of work. What you spend per message depends on the length and complexity of the conversation, the features and connected tools you use, model choice and effort level. Two people on the same tier can get very different amounts done, which is why no honest guide can tell you how many messages you get.

The limits are shared. Claude chat, Claude Code and Cowork all draw from the same allowance on one account, so a heavy afternoon in the terminal takes usage away from the chat window that evening. How Claude counts a usage limit covers the mechanics in detail.

What happens when you hit a Claude Max limit?

Unless you have turned on usage credits, Claude blocks new requests until the window resets. Nothing is billed. Claude Code is also rolling out a small, capped wrap-up allowance that lets a response already in progress briefly finish its current step after a five-hour limit; that usage counts toward your weekly limit. In Claude Code v2.1.234 or later, an interrupted task can wait in the open interactive session and continue automatically after the reset. In Claude Code the message names the ceiling you hit and the time it clears:

You've hit your session limit · resets 3:45pm
You've hit your weekly limit · resets Mon 12:00am
You've hit your Opus limit · resets 3:45pm
You've hit your Sonnet limit · resets 3:45pm

The session and weekly limits are shared across all models, so running /model and picking a cheaper model does not give you access back. The Opus and Sonnet limits are the exception, because each applies to that model family alone. Switch to a model outside that family with /model and you keep working on the same plan. If the notice on your screen says your Opus priority access is used up for the week, what that weekly Opus notice means covers when Opus comes back and whether usage credits or a bigger plan bring it back sooner. On Max and Ultra, the Fable weekly limit works the same way: switch to another model, or keep using Fable on usage credits. If you are locked out right now and not sure which of these you hit, work out which window you are waiting on before you do anything else, because the route out is different for each one.

Check Settings > Usage for a free way out before you pay. Anthropic occasionally offers eligible plans a limit reset, which sets either your five-hour session limit or your weekly limit back to full right away, depending on the offer. You use it from Settings > Usage on the web or in Claude Desktop, not from Claude Code in the terminal, and an unused reset expires on the date shown in the offer.

There is a paid escape. Usage credits are available on Pro, Max and Ultra, they are billed at standard API rates, and they apply to Claude chat and Claude Code alike. Run /usage-credits in Claude Code to open Settings > Usage, where you can set a monthly spend limit or turn on auto-reload. Credits are prepaid, so usage carries on until you reach the monthly spend limit you set or your balance runs out with auto-reload off. Separately, individual Pro, Max and Ultra subscribers can buy up to $2,000 a month of discounted usage bundles, which save up to 30% on standard credit rates and need usage credits turned on first.

Credits carry one detail worth knowing before you rely on them. Prompt caching lets Claude reuse the already-processed part of your conversation at a lower rate. In Claude Code, the main conversation defaults to a one-hour cache lifetime within your subscription allowance and five minutes on usage credits. Subagents and most other requests default to five minutes even within the subscription allowance. You can keep one-hour caching on usage credits by setting promptCacheTtl to 1h or CLAUDE_CODE_PROMPT_CACHE_TTL=1h, supported in Claude Code v2.1.242 or later. With the five-minute default, a break longer than five minutes expires the cache, so the next message processes the full conversation again. One-hour caching charges a higher rate for cache writes.

Why heavy Claude Code use changes the answer

A chat turn sends a question and gets an answer back. A Claude Code turn sends file contents, tool results and the whole conversation so far, then sends all of it again for each tool call inside the same task. That is why Anthropic's own guidance for admins says to budget more for a coding seat than for a chat seat. If you are the one setting that budget for a team rather than for yourself, what a Claude Code seat costs on Team and Enterprise puts the published seat prices next to per-token billing. If the coding agent is the only reason you are paying at all, what Claude Code costs on each plan works the same decision from that angle, with a worked monthly estimate. And if the prior question is whether to pay for the agent at all rather than how much, the hours it has to save to cover its own cost is the sum to run first, against your own hourly rate.

Four habits push usage up faster than anything you type. Leaving one session open all day means every request carries hours of history. Resuming a large session after a long break misses the cache, so the full context is processed from scratch. Scheduled tasks fire on their interval while you are away from the keyboard. Agent teams multiply everything: each teammate runs its own context window, and token usage scales with the number of active teammates and how long each one runs.

The practical result is a wide gap between a chat-only user and someone running a coding agent for most of the working day. Before you pay for the higher tier, spend a week on context discipline: /clear between unrelated tasks, /compact with an instruction about what to keep, and /context to see what is filling the window. Keeping Claude Code's context small goes through that properly, and choosing between Opus, Sonnet and Haiku matters just as much, since Opus has its own ceiling on top of the shared one.

How do you tell which Max tier you need?

1. Run a full billing month on Max. A busy week overstates your average, and the weekly ceiling is what you are testing.
2. Run /usage in Claude Code. It shows your plan usage bars and attributes recent usage to skills, subagents, plugins and individual MCP servers. Press d for the last 24 hours and w for the last 7 days.
3. Read the flags in that breakdown. It calls out behaviour such as long context or cache misses when one of them accounts for a tenth or more of recent usage, and those are fixable without spending more money.
4. Open Settings > Usage on claude.ai as well. The /usage attribution breakdown is computed from local session history on that machine, so a second laptop, the desktop app on another device, or claude.ai itself is missing from that breakdown. The plan usage bars show your account's remaining allowance across devices.
5. Upgrade if you hit the weekly limit before the week is half over and step 3 did not explain why. Stay where you are if you have never seen a session limit message.

Moving from Max to Ultra changes your bill and your ceilings, including removing the five-hour limit once your account is verified. You keep access to Claude Code and Cowork on the same account.

When Claude Max is the wrong answer

A subscription is a fixed price with a capped allowance. Past the cap you either wait, use a limit reset if you have been offered one, or pay extra at API rates through usage credits. That trade works well for one person working interactively, and it works badly for two other shapes of work.

The first is automation that must not stop. When a subscription allowance runs out, it waits for the reset unless usage credits are on, and a limit reset has to be used by hand in a browser or Claude Desktop. API billing charges for what you use instead: the key keeps serving as long as the account can pay. Most Console accounts run on prepaid credits, and once they run out the API stops answering until you add more, so turn on auto-reload for anything unattended. A nightly job that dies at 3:45pm on a session limit costs you more than the bill would have.

The second is bursty work. Two quiet weeks and one heavy week still cost a full month of subscription. Per-token billing follows the work instead of the calendar.

There is also an access question. A subscription belongs to one person's login, so an unattended agent on a server should hold its own API key. That keeps automated work off your interactive allowance, and it gives you spend reporting per key. Cost controls for AI agents running on a VPS covers the guard rails to set before you point a key at a server.

Here is the comparison against Anthropic's published figures for teams that pay per token.

[
  {
    "label": "Max subscription",
    "price_usd": 100
  },
  {
    "label": "Ultra subscription",
    "price_usd": 200
  },
  {
    "label": "API, low end of published range",
    "price_usd": 150
  },
  {
    "label": "API, high end of published range",
    "price_usd": 250
  }
]

Read those API rows carefully. They are averages Anthropic publishes for enterprise deployments, and not a quote for your workload. They land between $150 and $250 per developer per month, which means a heavy Claude Code user often pays more per token than Ultra costs. That is the argument for Ultra, and it is the reason the plan exists.

[
  {
    "label": "Average across deployments",
    "usd_daily": 13
  },
  {
    "label": "90% of users stay below",
    "usd_daily": 30
  }
]

To price your own usage properly, start from what a million Claude tokens costs, then compare the two billing models with the Claude API against a subscription. If more than one person needs this, seats change the arithmetic again, and Claude Enterprise plan pricing covers the seat model, central billing and admin controls.

FAQ

Is there a cheaper Claude Max tier?

Max costs $100 a month against $200 for Ultra, as of October 2026. They are separate plans; Ultra replaces Max 20x. Max gives 5 times Pro's usage every five hours and every week, while Ultra gives 10 times Pro's weekly usage and has no five-hour limit for verified accounts.

Does Max 20x unlock anything Max 5x does not?

Max 20x is now Ultra, while the $100 plan is Max. Both plans include Claude Code, Cowork and early access to new features and models. Ultra doubles Max's weekly allowance and removes the five-hour limit for verified accounts. Monthly API credits are $100 on Max and $200 on Ultra. Both cost $20 per Pro-sized weekly allowance, as of October 2026.

What happens when I hit my Claude Max usage limit?

Claude blocks new requests and tells you when the window resets, for example You've hit your session limit · resets 3:45pm. Claude Code is rolling out a small, capped allowance to finish the step already in progress after a five-hour limit, counting toward your weekly usage. The session and weekly limits cover every model, so switching model with /model does not restore access. An Opus or Sonnet limit is the exception, because each blocks only that model family, so switching to a model outside it keeps you working. If Anthropic has offered you a limit reset, you can use it from Settings > Usage to refill the session or weekly limit straight away. You can also buy usage credits with /usage-credits, billed at standard API rates and usable in both chat and Claude Code.

Do Claude Code and Claude chat share the same limits?

Yes. On Pro, Max and Ultra, usage is shared across Claude chat, Claude Code and Cowork on the same account. Pro and Max have a shared five-hour session allowance and weekly ceiling; verified Ultra accounts have no five-hour limit. Max and Ultra also have a separate Fable weekly limit within their shared allowance. Run /usage in Claude Code or open Settings > Usage on claude.ai to see what is left. The /usage attribution breakdown comes from local session history on that machine, so usage from another device is not counted in that breakdown; the plan usage bars cover your account across devices.

When should I pay per token instead of subscribing to Claude Max?

When the work must not stop, and when the work is bursty. Once the ceiling is reached, a subscription waits for the reset unless you pay for usage credits or use a limit reset by hand. An API key keeps serving and charges for what it uses as long as its prepaid credits last, so give it auto-reload. Automation and anything running unattended on a server belong on an API key for that reason. One person working interactively on most days is usually cheaper on a Max plan.