Claude Pro vs Max vs Ultra: Is 5x the Usage Worth 5x the Price?

Anthropic sells three prices for one chatbot. Twenty dollars, one hundred dollars, two hundred dollars. In late October the top rung got a new name: Max 20x is now Ultra, and the $100 plan is simply Max. The middle rung is still the one people write in to ask about, because $100 a month is where a subscription stops feeling like a subscription and starts feeling like a line item.

So here is the arithmetic that nobody puts at the top of these comparisons. Max costs five times what Pro costs and gives you five times Pro's usage. Ultra costs ten times what Pro costs and gives you ten times Pro's weekly usage. Every rung charges the rate you were already paying on monthly Pro, in a bigger block, with no discount for buying in bulk.

That is not a criticism of Claude, which remains very good. It is a description of what the transaction is. And once you see it, the question stops being "is Max better than Pro" and becomes something much easier to answer: do you need more of the thing you already have, and which of Claude's two clocks is the one stopping you?

The Short Answer: Who Should Move From Pro to Max or Ultra

If this is you	Verdict	Why
You hit the five-hour limit most working days	Yes, Max	Five times the session allowance is the problem Max is built to solve
You run out of your week, not your session	Max, or Ultra if that's not enough	Max now multiplies the week by five, Ultra by ten
You want to work for hours without a reset	Ultra	The only plan with no five-hour limit, once your account is verified
You want the Fable tier without a token meter running	Yes, Max	The one model row on the table that genuinely changes
You already spend $100 a month on Claude's API	Yes, Max	Max returns its price as API credits
You hit the limit once or twice a month	No, usage credits	$80 of credits goes a long way at Anthropic's own rates, further in a bundle
You want better answers than Pro gives you	No	Same models, same context window, no capability row changes
You are on annual Pro	Check your billing address first	The credit for your unused year depends on it matching

How Much Claude Max and Ultra Cost

Max is $100 a month. Ultra is $200 a month. Both are monthly only, with no annual option and no discount at any volume.

Getting those two numbers used to take a trip to the help centre, because the pricing card said "From $100" above a table with a Max 20x column and no price for it. The pricing page now shows two cards, Max at $100 and Ultra at $200, and the help centre's Ultra article explains the change in two sentences: "The Max 20x plan is now called the Ultra plan. Existing Max 20x subscribers move to Ultra automatically, with no price change." Anthropic's release notes don't date it. The pricing page still showed Max 5x and Max 20x when I read it on 20 October, and showed Ultra by 31 October. Both help articles add a caveat worth knowing before you subscribe on your phone: those are web prices, and "mobile pricing may vary depending on your app platform."

There are no discounts on either plan. Anthropic's Pro article says it doesn't "offer standard discounted pricing for individual plans (Pro, Max, and Ultra)," and its support team cannot issue one-off discounts or coupons.

The upgrade path has one trap in it. Moving up a tier is charged prorated, which is fine, and if you go from Max to Ultra with API credits claimed, you get a prorated top-up straight away. Moving from an annual Pro plan to Max or Ultra converts your remaining balance into account credit, which is generous. But that conversion has a condition attached: the billing address on your new subscription has to be the same as the one on your old Pro subscription. If it changed, and people move house, Anthropic's instruction is to cancel Pro, wait for the subscription end date, and then sign up again with the new address. That means eating the rest of a year you have already paid for.

Both plans also include API credits, which Anthropic's release notes date to 7 October. Max comes with $100 a month and Ultra with $200 a month for building on the Claude Platform, meaning the Claude API, the Playground in the Console, Claude Managed Agents and the Agent SDK. The help article is just as clear about what they do not pay for, and that list matters more to most people reading this. It excludes interactive Claude Code in the terminal, IDE, desktop app or web, extra usage in Claude, Claude Code or Cowork, and Claude through Amazon Bedrock, Google Cloud or Microsoft Foundry. Pro gets no credits at all.

The terms are tight. You claim the credits on claude.ai under Settings and Billing, after seven days on the plan, by linking one Claude Console organization, and you cannot change that organization yourself later. They refresh every billing cycle and do not roll over, so whatever you leave unspent expires. And they leave your usage limits in Claude where they were.

Five Times the Usage for Five Times the Money

Put the whole consumer ladder in one table and divide. A "unit" here is one Pro plan's weekly usage, which is the only unit Anthropic gives us.

Plan	Price per month	Weekly usage vs Pro	Cost per unit of Pro usage
Pro, annual	$16.67	1x	$16.67
Pro, monthly	$20	1x	$20.00
Max	$100	5x	$20.00
Ultra	$200	10x	$20.00

Read the right-hand column twice. Every monthly rung costs $20 a unit. Not a fraction less. The $100 plan is five Pro subscriptions, priced at retail, delivered to one login, and the $200 plan is ten.

And against annual Pro, which is what a lot of people upgrading from are actually on, Max and Ultra are twenty per cent more expensive per unit of Claude. You move up a tier and every message you send costs you more than it did last month. That is an unusual shape for a pricing ladder. Most products get cheaper per unit as you buy more, because that is the entire argument for buying more.

Before the rename, the $200 rung looked like the exception. Max 20x promised twenty times Pro's usage per five-hour session for ten times the price, which halved the rate on the one clock Anthropic gave a multiplier for. Ultra is priced at exactly Pro's rate for the week. What it sells instead of a discount is the removal of the short clock: verified Ultra accounts have no five-hour limit at all. That is the most interesting thing Anthropic has done to this ladder, and it gets its own section below.

There is one exception, and it only applies if you write code against Anthropic's API. Max's and Ultra's API credits are worth the plan price, so every dollar you would have spent on the API anyway comes off the bill. A developer already spending $100 a month on API calls gets Max's chat allowance for nothing extra, and one spending $200 gets Ultra the same way. Someone spending $50 pays an effective $50 for five times Pro, which is $10 a unit. If you have never opened the Claude Console, none of this applies to you. The credits expire unused each month, and the table above is your real price.

I want to be fair about one thing here. Nothing about this is dishonest. Anthropic publishes the multipliers and the prices, both in plain language, and the sum is not hidden. It just never appears as a rate, and rates are how people decide whether a bigger package is a bargain.

Anthropic's Comparison Table Has 56 Rows. Six Change.

The compare-features table on the pricing page runs across four sections. I counted the rows in each and checked every cell in the Pro, Max and Ultra columns against each other.

Section	Rows	Pro vs Max	Max vs Ultra
Features and capabilities	27	0	0
Models and usage	12	5	3
Security and administration	12	0	0
Payment options	5	1	0
Total	56	6	3

Zero, in the section that describes what Claude can do. Claude Code, Research, Projects, Connectors, Skills, Memory, Artifacts, voice mode, incognito chats, Claude in Chrome, the Microsoft 365 and Outlook integrations, Design, Slides, Docs, Dashboards, Science, scheduled tasks, file creation with code execution. Every one of those is identical on the $20 plan and the $200 plan. Twelve more rows under Security and administration are a clean column of No on all four consumer tiers, so they cannot differ either.

When I first counted, in September, the table had 49 rows and three of them differed. By 20 October it had 54 rows and four. The Ultra change added two rows under Models and usage, one for each of Claude's clocks, and both of them differ.

That leaves six cells in 56 rows:

The five-hour limit. Pro reads 1x, Max 5x, and Ultra "No 5-hour limit", with a footnote: "No 5-hour limit for verified accounts."

The weekly limit. Pro 1x, Max 5x, Ultra 10x. These two rows settle a question the old table left open, which I come back to below.

Fable. Pro reads "Usage credits." Max and Ultra read "50% of weekly limits." This is the real one for people who chat, and it gets its own section.

Monthly API credits. Pro reads No, Max "$100/month", Ultra "$200/month", under a NEW label. It is worth the full price to a developer and nothing to anyone who only chats, for the reasons above.

Priority access at high traffic times. Free No, Pro No, Max Yes, Ultra Yes. Anthropic's own help centre article on the Pro plan lists "priority access to Claude during high-traffic periods" as a Pro benefit, which directly contradicts the table. That contradiction was still live when I rechecked on 3 November, and I wrote it up in more detail when it first turned up.

Billing cycle. Pro is "Monthly and annual." Max and Ultra are "Monthly." This one runs the wrong way. The upgrade takes annual billing away from you, which is how both plans end up more expensive per unit than the plan below them.

Between Max and Ultra, three cells change: the two clocks and the API credits. Fable, priority access and billing are the same on both.

One row deserves a special mention for not changing: context window. It reads "Up to 1M, varies by model" on all four columns, and that includes Free. Whatever you are paying Anthropic, from nothing to $200, you get the same working memory. If you came here hoping Max would let Claude hold more of your document at once, it will not.

Anthropic Describes Max the Same Way It Describes Pro

Anthropic writes a "Key benefits" list at the top of each plan's help centre article. The Max article lists six. The Pro article lists five. I put them side by side, and the overlap is the most revealing thing on either page.

Max key benefit	Also a Pro key benefit?
More usage capacity	Pro says "more usage per session than the Free plan"
Fewer interruptions	Restates more usage
Priority access	Yes, Pro benefit two
Access to Claude Code	Yes, Pro benefit four, and the table marks it Yes on Pro
Longer, multi-step tasks	Yes, Pro benefit five, word for word
Monthly API credits	No. The Pro article says the plan does not include API usage through the Console

Two of Max's six benefits are the same fact written twice. Three more appear, in Anthropic's own words, on the Pro page. The sixth, monthly API credits, is the only Max benefit Anthropic does not also claim for the $20 plan, and it is aimed at developers rather than at people who chat.

The Ultra article has the same list with one swap. "Fewer interruptions" becomes "No five-hour limit for verified accounts," and that is the one benefit on any of the three pages that describes a different product rather than more of the same one.

The Max card on the pricing page used to carry a bullet that appeared nowhere else, "Higher output limits for all tasks," with no number behind it anywhere on Anthropic's site. It is gone now. In its place is "Early access to advanced Claude features," which the Pro article also promises its own subscribers as "early access to new features."

None of this makes Max a bad plan. It makes Max a quantity plan, sold with the vocabulary of a quality plan, which is why write-ups still circulate claiming Max unlocks the top model. It does not. Opus is marked Yes on Pro.

The One Row That Genuinely Changes: Fable

Anthropic's model lineup has a tier above Opus called Fable, and this is where the $80 actually goes for anyone who only chats.

On Pro, the Fable cell says "Usage credits." That means Fable is not part of your plan's allowance at all. You turn on usage credits, you prepay, and Fable bills at standard API rates from your first message. On Max and Ultra, Fable is included, drawing from your regular weekly allowance up to half of it, and using it up faster than other Claude models do.

Here are the rates that cell is pointing at, from Anthropic's own API pricing.

Model	Input per million tokens	Output per million tokens
Fable 5.1	$10	$50
Opus 5.5	$4	$20
Sonnet 5.5	$2	$10
Haiku 5.5	$0.10	$0.50

Haiku's rate is for prompts up to 100,000 tokens. Above that it is $0.50 and $2.50.

Now you can price the upgrade properly, because the gap between Pro and Max is $80 a month and usage credits are the direct alternative to paying it. Max's API credits do not enter this sum. Anthropic excludes extra usage in Claude, Claude Code and Cowork from what they cover, so they cannot pay for Fable in a chat, and Pro does not get them anyway.

Eighty dollars at Fable rates buys roughly 8 million input tokens, or 1.6 million output tokens, or any mix in between. That upper figure is somewhere around a million words of Claude's writing. For someone whose Fable use is a few genuinely hard problems a week, $80 of credits is not a near thing. It is not close.

And the credits got cheaper. Anthropic now sells them in prepaid bundles on Pro, Max and Ultra, at up to 30% off: $45 for $50 of credit, $200 for $250, $700 for $1,000. Bundle credit only kicks in once you are past your plan's own limits, which is exactly where Fable on Pro lives.

The honest qualifier: agentic runs and long conversations burn input tokens far faster than chatting does, because the whole thread gets re-sent with every turn, and reasoning models bill their thinking as output. So treat that as a ceiling rather than a promise. But it does mean the upgrade has a real test attached to it. Turn on credits for a month, put your hard problems on Fable, and look at the bill. If it comes in under $80, Max was never the cheaper option. If it comes in at $150, Max just paid for itself and you should switch.

There is also a model above Fable that no consumer plan reaches. Mythos 5.1, which Anthropic calls its most capable model class for cybersecurity and biology research, goes only to vetted organisations through its verification programmes. Anthropic's model page says Fable 5.1 "is the same underlying model as Claude Mythos 5.1 with safeguards for cybersecurity and biology," so the Fable row on Max is as close as an individual subscriber gets. The differences between the three tiers you will actually pick from matter more day to day than either.

The Multiplier Now Covers Both of Claude's Clocks

This is the detail that decides whether Max fixes your problem, and it changed with the Ultra rename.

Claude meters you on two clocks. A session limit that resets every five hours, and a weekly limit that applies across all models. Until late October, Anthropic stated the Max multiplier "per 5-hour session" every time, in the pricing FAQ and the help centre alike, and published no multiplier for the week. So Max promised longer stretches without promising more Claude per month.

The pricing page now settles it. The Max card reads "5x Pro usage every 5 hours and every week," and the comparison table has a row for each clock with 5x in both. Max buys you both things now: fewer cut-offs in the middle of an afternoon, and five Pro weeks in one. Anthropic keeps the small print that says it may limit usage "in other ways, such as weekly and monthly caps or model and feature usage, at our discretion," so the week is not unlimited. But it is a number.

Ultra goes the other way. It multiplies only the week, by ten, and deletes the short clock for verified accounts. Anthropic verifies Ultra accounts automatically, "using factors like your account and payment history," and until it has, "a temporary limit applies." If Settings shows no Current session bar, you are verified.

That maps onto two different complaints. If your problem is "I get cut off in the middle of a long afternoon and lose my thread," Max fixes it and Ultra removes it. If your problem is "I run out by Thursday every week," Max gives you five times the week and Ultra ten. The catch with Ultra is the flip side of its selling point: with nothing stopping you after five hours, a heavy day can spend a large share of the week. The API credits do not help with either clock, since Anthropic states they leave your usage limits in Claude, Claude Code and Cowork unchanged. How both clocks behave, what the error messages mean and when they reset is worth ten minutes before you spend $1,200 a year.

Work out which one is stopping you first. Settings has a Usage section with progress bars for the current session and the week, which is genuinely more than most vendors show you, and it will tell you the answer in about four seconds.

Claude Max vs ChatGPT Pro: The Same Prices, the Same Credits

I read OpenAI's pricing page and help centre on 3 November, and the two ladders have moved closer since September.

	OpenAI	Anthropic
Entry paid plan	Go, $6	none
The $20 plan	Plus, $20	Pro, $20
$100	Pro 100	Max
$200	Pro 200	Ultra
$500	Pro 500	none
Monthly API credits	The plan price, since 28 October	The plan price, since 7 October

Anthropic put API credits on Max first. OpenAI's release notes add the same benefit to Pro 100, Pro 200 and Pro 500 three weeks later: $100, $200 or $500 a month, claimed after seven days into one linked API organisation, and kept apart from ChatGPT's own usage allowance. Same shape, same amounts, same exclusion of the chat product itself.

Where the ladders differ now is the multiplier. Anthropic prints one on every rung. OpenAI's Pro article says only that "Pro 200 includes more usage than Pro 100" and that Pro 500 has the most. The version I read on 3 November had also doubled the Work and Codex allowance on Pro 100 and Pro 500 and cancelled the cut to Pro 200 that OpenAI announced on 29 September. Pro 500 is the only rung either company sells with no counterpart, and the only personal plan with Ultrafast, OpenAI's faster mode in ChatGPT Work and Codex. I went through what the $500 buys separately.

That convergence tells you something the single-vendor comparison cannot. The $100 tier is not Anthropic's judgment about what five times Claude is worth. It is the industry's standard price for selling headroom, and since October it comes with the same API credits on both ladders. If you find yourself deciding between $100 of Claude and $100 of ChatGPT, neither purchase is the good deal in the room, and the cheaper comparison is one rung down where both companies compete hard for the same $20.

Two differences matter. OpenAI puts a $6 plan on its ladder and Anthropic does not, so Claude's floor for paying anything at all is $20. And OpenAI sells a $500 rung that Anthropic does not.

Is Claude Max Worth It? Verdicts by How You Work

Writers working on book-length projects. Probably yes, on Max. Long-document work is where Claude earns its reputation, and it is also the fastest way to spend a session allowance, because every turn re-sends the document. If you are editing a manuscript and getting stopped mid-chapter, that is the exact complaint Max fixes.

Researchers. Yes if Research runs are your daily habit, no otherwise. Research mode is on Pro already, and it consumes an allowance quickly enough that heavy users find Pro's ceiling in the first week. This is the group most likely to actually want Max, and the one most likely to outgrow it.

People running long agent tasks or Claude Code all day. This is who Ultra is for. A task that runs through the afternoon no longer stops at a five-hour reset. Keep one eye on the weekly bar, because ten Pro weeks spent at that pace go faster than the number suggests.

People who want the Fable tier. Do the credits test first. Most of you will find $80 of credits covers a month of hard problems with room left over, a bundle makes it cheaper still, and you can stop paying in any month you do not need it, which Max does not let you do.

Developers already paying for the Claude API. Yes. If your API bill is $100 a month or more, Max's credits cover it and the chat plan costs you nothing extra, and a $200 bill does the same for Ultra. Two catches. The credits go to one Console organization that you pick once, and they do not cover interactive Claude Code, so a Claude Code habit still runs on your plan's limits.

Consultants and analysts billing by the hour. Yes, and the maths is not close. If waiting for a reset costs you one billable hour a month, Max has paid for itself. This is the cleanest case on the list and it has nothing to do with features.

Anyone whose complaint is answer quality. No. Not one capability row changes. Claude generates no images on any plan, the context window is identical from Free upward, and the model list is the same apart from Fable. Paying Anthropic more money gets you more Anthropic.

Anyone currently on annual Pro who is merely curious. No, or at least not this month. You would be trading a $16.67 rate for a $20 rate, giving up annual billing permanently, and relying on your billing address not having changed to get your unused balance back as credit.

What Else $80 a Month Buys

Here is the thing about Max being priced as five Pro subscriptions: it is worth asking whether five of the same is what you actually wanted.

Priced separately, the major assistants run to roughly $110 a month all in. At $100, Max costs nearly the whole industry and gives you one vendor's opinion, five times over. That is a fine trade if your bottleneck is genuinely volume. It is a poor one if you have ever asked a model something and thought, I wonder what a different model would say about this.

izzedo chat is built for the second case. It puts Claude's tiers in one picker beside the flagships from OpenAI, Google, xAI, DeepSeek, Moonshot and the rest, and the whole thing starts at $6 a month.

The specific comparison worth making, given everything above: the Fable tier that Anthropic meters at API rates on the $20 plan, and caps at half your week on the $100 and $200 plans, is a row in that list on the $6 plan. Claude Haiku is on the free plan with no card.

I want to be precise about the limits, because this article has spent two thousand words reading other people's fine print and it would be poor form to hide my own. There are two clocks here too, a five-hour one and a weekly one, and they live behind the same icon as the context meter rather than three menus deep in a settings page.

It is one allowance, metered by what each model costs to run, so heavy models spend it faster than light ones. It is fair use rather than unlimited, and the numbers live on the pricing page rather than in a blog post. Usage-based billing exists if you want to keep working past the allowance, the same shape as Anthropic's usage credits, and it stays off unless you switch it on.

What the ladder above cannot sell you at any rung is a second opinion. When Claude hedges on something that matters, you ask the same question of a model from a different lab in the same thread and compare the two answers, which is the habit I argued for at length in stop using one AI model. Ultra will not do that for $200. Neither will ChatGPT Pro.

The Bottom Line

Claude Max is worth it when Pro's limits are costing you working hours, and since the rename it multiplies both clocks, which makes it a cleaner purchase than it was. Ultra is worth it when the five-hour reset is the thing in your way and you are willing to watch the week instead. Both are worth it if you already pay Anthropic that much for the API. For everyone else there is almost no other reason.

The extra $80 buys six cells on a 56-row table. Two are the clocks, five times each. One is Fable, which you can price yourself against usage credits in a single month. One is $100 a month of API credits, worth the full price to a developer and nothing to anyone who only chats. One is priority access during busy periods, which Anthropic's own Pro article also promises to Pro subscribers. And one is the loss of annual billing, which is how a plan five rungs up ends up charging more per unit of Claude than the plan you are on.

What surprised me, after a day inside Anthropic's pricing pages and help centre, is how consistent the company is about all of it. Every number I used here came from Anthropic. It now states the multiplier on both clocks. It tells you there is no fixed message count. It names the discretionary monthly cap in its own small print. The marketing is not misleading so much as arithmetically quiet: every monthly rung costs $20 a unit, and nowhere on the page does it say so.

So run the test that costs nothing. Open Settings, look at the Usage bars, and find out whether it is the five-hour clock or the weekly one that keeps stopping you. If it is the session, Max fixes it and Ultra removes it. If it is the week, Max buys five Pro weeks and Ultra ten, at the rate you are already paying. And if what you actually want is a second opinion, no rung on either company's ladder sells one.

izzedo's free plan has Claude in it, takes no card, and lets you put the same question to three labs before committing $1,200 a year to one of them.