Anthropic released Claude Fable 5.2 on November 9, 2026, and put the model inside the weekly usage limits of Claude Pro at no extra cost for one week. The launch offers page ends that window on November 16, 2026 at 11:59 pm Pacific. Two documents describe it, and the shorter one is the announcement.
What was actually released
Fable 5.2 is a model update, but the offer shipped with it is written plan by plan, and it grants less than “extra Fable usage” suggests.
On Pro plans and standard seats on Team plans, all Fable models sit inside the weekly usage limits from November 9 through November 16, up to 50% of those limits, at no extra cost, with one Fable reset. Max, Ultra and premium seats get no included week but one to three resets, counted by subscription length.
| Plan or seat | What the launch offer gives |
|---|---|
| Pro, standard seat on Team | All Fable models inside the weekly limits, Nov 9 to Nov 16, up to 50% of the limit, plus one reset |
| Max, Ultra, premium seats | One to three resets by subscription length, no included week |
| Free, standard seats on legacy Enterprise, self-serve Enterprise, HIPAA-ready Enterprise | Nothing |
5.2, 5.1 and 5 share one allowance and one set of resets.
A reset is not extra capacity: it gives back usage already spent on Fable models, and on accounts with a five-hour limit it refills that whole window, including usage on other models. The count is mechanical: under six months of paid billing, one reset, six to 12 months, two, more than 12 months, three. Eligibility needs paid billing in place before November 9, 2026 at 10:00 am Pacific.
The Pro reset dies with the week, on November 16 at 11:59 pm Pacific. Resets on Max, Ultra and premium seats run to March 8, 2027, then lapse.
The model ships at $8 per million input tokens and $40 output, with a fast mode at twice the rate, $16 and $80.
What the claim rests on
The offer is documented twice, and the two texts diverge.
The announcement gives it one paragraph: the window is “this week”, the resets are “one to three” by subscription length, and the 50% figure arrives with no start and no end date. Nothing there names November 16, the 10:00 am cutoff, the March 8, 2027 expiry, or the excluded plans.
The announcement gives the shape of the offer; the dates and the exclusions live only on the help page it links to.
Those details come from the launch offers help page, read on November 10, 2026. A reader who stops at the announcement has no deadline in hand, and the deadline is the news.
What neither page quantifies is what the week is worth. Pro subscribers are told Fable can take half the weekly limit, and that Fable models consume limits faster than other Claude models, with no multiplier attached. How many tasks that buys depends on prompts, file sizes and model choice, none of it published. Nor does Anthropic say what the same week costs on usage credits, which is the number a buyer wants.
The speed claims have the usual shape: up to 30% faster than Fable 5.1, a fast mode up to 2.5 times quicker. Both are the company’s own runs, with no independent replication at launch.
| Model | Input, per million | Output, per million | Cache read | Cache write |
|---|---|---|---|---|
| Claude Fable 5.2 | $8 | $40 | $0.20 | $10 |
| Claude Fable 5.1 | $10 | $50 | $0.25 | $12.50 |
| Claude Opus 5.5 | $4 | $20 | $0.20 | $5 |
| Claude Haiku 5.5 | $0.10 | $0.50 | $0.01 | $0.125 |
What the pages do not say
No token or dollar figure behind the 50% cap, no multiplier for how fast Fable burns a weekly limit, and no independent measurement of the 30% claim. Fast mode trades 2.5 times the speed for twice the token price.
On this page
What it changes for Pro and Max subscribers
The default model does not change. Fable has to be picked in the model picker, and it then draws on the same weekly allowance as Opus 5.5 and Sonnet 5.5, capped at half of it. A week mostly on Fable leaves half the allowance for everything else.
The reset is a deadline inside a deadline. On Pro it expires on November 16 at 11:59 pm Pacific with the included week, cannot be undone once used, and is lost on cancellation or downgrade. On Max, Ultra and premium seats it lasts to March 8, 2027, which makes it a reserve.
After the window, Fable goes back to usage credits, and the plan page read today says so: Pro lists Fable as paid with usage credits, Max as included up to half the weekly limit. The permanent structure of the plans did not change on November 9.
That asymmetry is what to carry into the week. The allowance is the cheapest way to test Fable on real work, because outside the window the same tokens are billed to credits. But Fable does not stay in the limit, so the week answers whether Fable is worth $8 and $40 to this workflow, not whether it is good.
The Neural Minute take
The number that matters is not 30% and it is not 5.2. It is $8 and $40.
A reset doesn’t raise your limits. It restores usage so you can keep working within your existing limits.
That sentence is the offer in one line: a restore, not a grant. Anthropic hands Pro subscribers a week of a model listing at $8 per million input tokens and $40 output, which is 80 times the input rate of Haiku 5.5 at $0.10 and 80 times its output rate at $0.50. A model at 80 times the price of the small one is a tool for the task the cheap model gets wrong.
Anthropic says as much, recommending Opus 5.5 as the starting point for most coding work and keeping Fable for the most ambitious tasks. With a one-week window, that advice describes a demo.
Which makes the deadline the useful part. A Pro subscriber has until November 16 to put Fable on work Opus 5.5 gets wrong, compare both on the same task, then price the result at $8 and $40. If the week replaces a paid API test, it has paid for itself. If Fable becomes the default, the bill moves to credits or to Max, the plan the rate card points at.
Two reservations. Nobody can say how far half a weekly allowance goes on Fable, because no multiplier is published, so the honest measure is the account’s own usage page. And the 30% claim stays a vendor figure until a second party reproduces it. A plan change is read on the invoice, and the part of this announcement that reaches an invoice is the date.
Sources
- Introducing Claude Fable 5.2 and Claude Mythos 5.2, anthropic.com, November 9, 2026: https://www.anthropic.com/claude-fable-and-mythos-5-2
- Anthropic’s launch offers help page, linked from the announcement, read on November 10, 2026: the November 9 to November 16 window, the 50% cap, the resets by subscription length, the March 8, 2027 expiry, the billing cutoff, the excluded plans, and the sentence quoted above
- Anthropic’s rate card, read on November 10, 2026: Fable 5.2 at $8 and $40 per million tokens, fast mode at twice that, Fable 5.1 at $10 and $50, Opus 5.5 at $4 and $20, Haiku 5.5 at $0.10 and $0.50
- The figures in both tables are Anthropic’s; the multiples between them are our arithmetic