Notes on coding agents and the pages they leave behind, on moving between notes apps, and on the plain mechanics of keeping what you think. Written and published from a Sumibako vault.
Claude pricing in November 2026: every plan and model, after Ultra replaced Max 20x
Claude pricing for November 2026: Free, Pro, Max, the new Ultra plan, Team, Enterprise and every API model, with what changed and which plan to buy.
Prashant Shaw
Founder, Sumibako
8 min read
On 23 October I wrote about Claude's usage limits and printed a table of the plans. A week later one of the plans in that table did not exist any more. Max 20x is now called Ultra, it has lost its five-hour limit and gained a weekly multiple, and the Max plan has quietly become a plan that comes with a hundred dollars of API money. That is three changes in the price of one blog post, which is roughly Anthropic's release cadence.
So this is the whole price list in one place: every plan, every current model on the API, what changed since September, and which one to buy. Everything was checked on 3 November 2026 against Anthropic's pricing page, its help centre and its API documentation. Prices are in US dollars, before tax.
💳 The plans, on one card
Plan
Price
Usage, from the pricing page
Claude Code
Free
$0
Limited, every 5 hours and every week
No
Pro
$20 a month, or $17 a month billed yearly ($200 up front)
10x Pro every week, no 5-hour limit for verified accounts
Yes
Team, standard seat
$25 a seat a month, or $20 billed yearly
More than Pro
Yes
Team, premium seat
$125 a seat a month, or $100 billed yearly
5x a standard seat
Yes
Enterprise
$20 a seat a month, billed yearly, 20 seats minimum, plus usage at API rates
Set by your admins
Yes
Max and Ultra are monthly only. Team is for 2 to 150 people, and you can mix seat types. The one thing every plan shares is the sentence underneath: "Usage limits apply." Anthropic still publishes no message count, only multiples, and the pricing page explains why in one line: "there's no fixed message count."
🔄 What changed since September
The plans and the models both moved this autumn, in roughly this order, according to Anthropic's release notes:
1 September: Claude Fable 5.1 launched. On Pro it costs extra (more on that below).
22 September:Claude Opus 5.5. "Input and output tokens are $4 and $20 per million, 20% less than Opus 5." Cache reads dropped 60%, and Anthropic says that at default settings it "will cost 40% less than Opus 5 on typical workloads". The same post raised five-hour limits on Pro, Max, Team and seat-based Enterprise.
28 September:Claude Sonnet 5.5, at the same $2 and $10 as Sonnet 5, with cache reads at half the price.
7 October:Claude Haiku 5.5, which "now costs around 75% less to run" than Haiku 4.5, and monthly API credits for Max and Team subscribers.
End of October: "The Max 20x plan is now called the Ultra plan. Existing Max 20x subscribers move to Ultra automatically, with no price change."
The direction is clear and pleasant: no model got dearer, cache reads got much cheaper on every family, and the subscriptions grew API credits. The catch is in the details, which is where the rest of this post lives.
🏔️ Ultra, and what "no 5-hour limit" means
Anthropic's Ultra article has the facts. Ultra is $200 a month and "includes 10 times the Pro plan's usage every week." Verified accounts have no five-hour limit at all; the weekly limit is the only wall.
Two details are worth reading twice. First, being verified is not something you do. Anthropic verifies accounts automatically "using factors like your account and payment history", and until it has, "a temporary limit applies." Typing your phone number at checkout does not count. To find out where you stand, open Settings, then Usage: "If you don't see a Current session bar, your account is verified."
Second, the multiple changed what it measures. The old pricing page sold Max 20x as twenty times Pro per five-hour session, as I quoted in October. Ultra is sold as ten times Pro per week. Anthropic never published a weekly multiple for Max 20x, so whether ten a week is more or less than before is a question its own pages cannot answer, and I would not trust anyone who answers it with confidence. What is certain is the shape: if you used to hit the five-hour wall in the middle of a long Claude Code run, that wall is gone, and the weekly one is the budget you manage now.
📝NOTE
Ultra still has "a separate weekly limit for Fable", and the same fine print as every plan: Anthropic "may limit your usage in other ways, such as weekly and monthly caps or model and feature usage, at our discretion."
🐉 Fable: the model that costs extra on Pro
Claude Fable 5.1 is Anthropic's most capable model in the apps, and it is the one place where the plan you pay for changes what a model costs. The Fable article is blunt about it: "On Pro plans, standard seats on Team plans, and standard seats on legacy seat-based Enterprise plans, both models run on pay-as-you-go usage credits."
That means a Pro subscriber pays $20 for the plan and then pays for every Fable message on top, at API rates, from the first one. On Max, Ultra and premium Team seats, Fable is included up to half the weekly limit. It is a share, not a bonus: asked whether Max users get 50% more for Fable, the article answers "No. You can use up to 50% of your weekly limit on Fable models, but your use of other models draws from the same usage limits".
Usage credits are the general version of the same idea. Turn them on in Settings, set a monthly cap, prepay, and when you hit any limit your requests "switch to pay-as-you-go pricing on usage credits and keep working", billed "at standard API pricing rates", per the usage credits article. It is the only plan feature that can make your bill bigger than the number on the pricing page, so set the cap before you need it, not after.
🎁 Max now comes with API money
Since 7 October, Max and Team, and now Ultra, include monthly credits for the Claude API. The amounts are in Anthropic's API credits page: $100 on Max, $200 on Ultra, and on Team $20 per standard seat and $100 per premium seat, pooled and "capped at $500 USD".
Read the summary table before you get excited:
Covers
The Claude API, Managed Agents, the Agent SDK and the playground
Doesn't cover
"Claude Code, extra usage in the Claude apps"
Expires
At the end of each billing cycle, with no rollover
Waiting period
"New subscribers can claim credits after 7 days on an eligible plan."
Not eligible
"Free, Pro, and Enterprise plans aren't eligible."
So the credit is for building things: an app, an agent, a script that calls the API. It will not pay for usage credits when you run out of Claude Code. You claim it by linking a Claude Console organisation in Settings, and you cannot change which organisation later yourself. For anyone already spending a hundred dollars a month on the API who also uses Claude Code all day, Max has become roughly free, which is either generous or the most elaborate customer acquisition scheme in software.
🔌 API prices, per million tokens
Model
Input
Output
Cache read
Best for, in Anthropic's words
Fable 5.1
$10
$50
$0.25
"demanding reasoning and long-horizon agentic work"
1️⃣ Batch halves input and output for work that can wait.
2️⃣ Caching costs 1.25x the input price to write for five minutes, 2x for an hour, and much less to read, so anything you send twice should be cached.
3️⃣ Long context costs nothing extra on every current model except Haiku 5.5: "A 900k-token request is billed at the same per-token rate as a 9k-token request." Haiku 5.5 charges five times more past 100,000 tokens, which is why it has two rows.
4️⃣ US-only inference adds 10%, and fast mode on Opus 5.5 is $8 and $40 for up to 2.5x the speed.
One warning for anyone comparing with last year's prices: Claude 4.7 and later models use a newer tokenizer that "produces approximately 30% more tokens for the same text." A cheaper price per token on more tokens is still cheaper here, but by less than the headline.
🧮 Which one to buy
1️⃣ Free, to find out whether you like Claude. No Claude Code, and Projects are capped at five.
2️⃣ Pro at $20, if you use Claude Code for part of the day and chat for the rest. Pay yearly if you are sure; it is $17 a month that way. Budget separately for Fable, or use Opus 5.5, which Anthropic says "performs at the level of Claude Fable 5.1 on most work".
3️⃣ Max at $100, if you hit Pro's limits most weeks, or if you were going to spend money on the API anyway.
4️⃣ Ultra at $200, if the five-hour wall is what stops you, and you run long agent sessions that should not pause for lunch.
5️⃣ The API, if you are building a product. Usage credits on a plan are billed at the same rates, so the API is not a cheaper way to chat; it is a different thing.
Whatever you pick, the cheapest habit is the one in Anthropic's own advice: "Begin fresh chats for new topics to minimize the size of your context window." A long conversation is re-read on every message, and that is what drains a limit. The trick is carrying the useful part across, which is what I built Sumibako for: Claude writes the state of a piece of work into a page and the next chat reads it back, through a custom connector that works on every Claude plan, Free included. It is a notes app, not a discount, and it will not make Fable free on Pro. Nothing will, as far as I can tell, except Max.
The pricing page ends with "Price and plans are subject to change at Anthropic's discretion." Given the last six weeks, I would check it again before you finish reading this sentence.
