Anthropic released Claude Fable 5.2 on 9 November 2026, the new version of the strongest model it opens to every user. The price came down 20%. A million units of input cost ten dollars before and now cost eight. A million units of output cost 50 dollars and now cost 40. One billing unit is called a token, and a million tokens run to roughly 1,500 pages of text.
Anthropic says the new model produces answers up to 30% faster than Fable 5.1. It is available from today on the Claude website and apps, through the programming interface, and through the Amazon, Google and Microsoft clouds. Anyone paying by usage sees the cut land straight on the monthly bill.
Paid plans get extra usage this week
This week Anthropic is giving paid subscribers extra Fable usage. Every plan that includes the model gets one to three resets of its limit, depending on how long the subscriber has been paying. On the Pro plan and on a standard Team seat, Fable can take up to half of the weekly usage limit.
For a business on a monthly Claude plan, this is a good week to try Fable on real work: a long contract to check, a report to take apart, a messy spreadsheet. By the end of the week it will be clear whether the gap over the regular model is worth the share of the limit it eats.
Anthropic itself points to a cheaper model first
In the release note Anthropic recommends starting most coding and office work on Claude Opus 5.5, and moving up to Fable only when that falls short. Opus 5.5 costs four dollars per million input tokens and 20 for output, half of Fable's price even after the cut. When it came out in September, its price fell 20% in the same way. For simpler jobs, such as sorting email or a short reply to a customer, Sonnet 5.5 and Haiku 5.5 cost less again.
So a business that built a bot to answer customers has no reason to move it to Fable because of this release. The cut matters mostly to anyone already running Fable 5.1 on heavy tasks, or anyone who passed on it because of the price.
Its coding score passed Opus 5.5
On FrontierCode 2.0, a test of writing code unaided across a whole task, Fable 5.2 scored 49.1%. The previous version scored 32.4%, and Opus 5.5 scored 45.5%. Fable 5.2's figure was measured at the highest effort setting. One step below it scored 45.3%, almost level with Opus 5.5, and by Anthropic's account at about half the cost per task.
Anthropic measured all of these numbers itself, on tests it chose. Anyone paying by usage will get a truer picture from running a week of their own work and reading the bill.
One more change shows up in daily use: in early testing Fable 5.2 blocked about 55% fewer requests than Fable 5.1. The blocks target questions on cybersecurity and biology, and a blocked request is passed to an older model of the company's.
Fast mode costs double
Alongside the new version comes a fast mode that generates text up to 2.5 times quicker. It costs 16 dollars per million input tokens and 80 for output, twice the regular price. It is available in Claude Code, on Anthropic's platform, on Google's cloud and on Microsoft's service. For most businesses that speed is not worth twice the price. It earns its keep in a bot that has to answer a customer's complex question in real time.
On the same day Anthropic released Claude Mythos 5.2, the same model without the limits on cybersecurity and biology. It is open only to organisations that passed one of the company's verification programmes, and a small business will not get access.