Claude Fable 5.2: What's New, Pricing, and Who Should Use It

Claude Fable 5.2, released by Anthropic on November 9, 2026, costs less per token than its predecessor, but that does not guarantee a cheaper finished task. For small businesses, the immediate opportunity is temporary access on Pro plans and standard Team seats through November 16. Anthropic also released Claude Mythos 5.2, the same underlying model without certain safeguards, for approved organizations only.

Claude Fable 5.2 price and access at a glance

• API price per million tokens: $8 input and $40 output, down from $10 and $50 on Fable 5.1 and double the rate of Claude Opus 5.5
• Max ($100 a month), Ultra ($200), and premium Team seats ($125, or $100 billed yearly): included, up to half of the weekly usage limit
• Pro ($20, or $17 billed yearly) and standard Team seats ($25, or $20 billed yearly): usage credits only, charged separately at API rates, outside the launch offer
• Free plan: not available

Plan prices are per month or per seat, before tax, from Anthropic's pricing page.

How to try it before November 16

Pro plans and standard Team seats can use Fable 5.2 for up to half of their weekly limit through November 16, 2026, at 11:59 pm PT. Choose it from the model menu next to the send button, because your default model does not change. Fable uses limits faster than other Claude models, and Settings > Usage shows how much is left. Once the allowance or the week runs out, further use is charged to usage credits if they are turned on, so check that setting first. Max, Ultra, and premium Team seats get limit resets instead. The full offer terms cover who qualifies and when resets expire.

What is new in Claude Fable 5.2?

1. A lower token price, and a bill that depends on the setting

The API rate fell 20% from Fable 5.1. What a finished task costs also depends on effort, the setting in the model menu that controls how thoroughly Claude works. Claude apps start at Medium, and higher settings use up plan limits faster. In independent tests by Artificial Analysis, checked on November 10, 2026, an average task at Medium cost $3.80 on Fable 5.2, down from $5.56 on Fable 5.1. At Max, the top setting, it cost $26.54, about twice Fable 5.1's $12.99. These are averages from a lab test suite, and your own bill will depend on your tasks.

2. Faster writing

Anthropic says Fable 5.2 generates output up to 30% faster than Fable 5.1. Time to a finished task improved less in Artificial Analysis's runs: 8.3 minutes on average at Medium, against 9.8 on Fable 5.1.

3. Higher scores than Opus 5.5, at a higher cost

Anthropic suggests starting most work on Claude Opus 5.5 and moving up only when the job needs it. On Artificial Analysis's general index, Fable 5.2 outscored Opus 5.5 at every setting, 56 to 48 at Medium, while costing more per task at that setting, $3.80 against $2.45.

One result points the other way for paperwork. On an independent test built on long professional PDFs, Opus 5.5 scored higher at four of the five settings. That test does not use small-business documents, so it cannot say which model will read your contract better. It is a reason to compare the two on your own documents before paying more.

What Claude Fable 5.2 changes for a small business

The advice below comes from Anthropic's published material and independent lab results. Fable 5.2 was not tested hands-on for this article.

Long contracts and statements

This is the most reasonable place to try it. Anthropic lists analysis carried through to a finished document among typical uses, and Box, an early tester, says the model flagged salary deposits that did not match a bank statement in a loan file. That is one example from a launch announcement.

Pick two or three documents where you already know what is wrong, ask Fable 5.2 and your current model the same question, and see which one finds the problem. After the launch window, a document that reads 60,000 tokens and writes 3,000 costs about 60 cents in usage credits, plus whatever the model's thinking adds. A lawyer or accountant should still review anything that binds the business.

Quotes, invoices, and totals

Let the model read the job details and draft the wording, and let a spreadsheet or invoicing tool calculate the total. In a test Anthropic quotes from Oracle, Fable 5.2 returned the expected totals while two other models produced sums 5 to 8 times too high. A single correct run is not enough reason to trust any model with the arithmetic.

Customer replies and social posts

Neither task appears in Anthropic's benchmark table. At usage-credit rates, a short reply (3,000 tokens read, 300 written) costs about 3.6 cents on Fable 5.2 and 0.9 cents on Sonnet 5.5, before the extra tokens from thinking, so moving this work up would need a visible improvement to justify it. A reply also needs your prices, services, and tone, whichever model writes it.

Empacer is a voice-first AI business assistant for small service business owners. It keeps your business details on hand and prepares posts, customer replies, and documents for you to review.

Who should try it now and who can wait

• Pro or a standard Team seat, with no document task that keeps going wrong: skip it. Your plan and default model stay the same after November 16.
• Max, Ultra, or a premium Team seat: Fable 5.2 is already included, so the decision is how to spend it. It is capped at half of the weekly limit and draws it down faster, which makes it better saved for the hardest documents.
• Free plan: a paid plan is worth considering only if Haiku 5.5 and Sonnet 5.5, the models included on Free, keep failing a task you care about.
• A tool or automation built on Claude: the vendor chooses the model. The lower rate may reduce its costs, and whether that changes your price is its decision.

FAQ

Is Claude Mythos 5.2 available to small businesses?

Only to organizations approved through Anthropic's Cyber or Life Sciences Verification Programs. Fable 5.2 is the public version. It blocks high-risk cybersecurity work and most research biology, and a blocked question is answered by an older Opus model.

Why can't I see Fable 5.2 in the model menu?

Fable models need a paid plan. On Team plans, an owner controls which models are available. Also check under "More models."

Does Anthropic keep what I paste into Fable 5.2?

It retains prompts and responses on Fable models for at least 30 days, then deletes them automatically unless a safety investigation or legal requirement applies. Model training is listed as opt-out on individual plans and off by default on Team, so check that setting before pasting client details.