If you pay for Claude by the token, Fable 5.2 just got 20% cheaper
Anthropic released Claude Fable 5.2 on 9 November alongside Mythos 5.2, a restricted version. The API price drops to $8 per million input tokens and $40 per million output, from $10 and $50 for Fable 5.1, and its robot-control score more than doubles.
Anthropic has a new model, and it costs less than the last one. Claude Fable 5.2 launched on Monday 9 November, and the official model page puts the API price at $8 per million input tokens and $40 per million output tokens. Fable 5.1, released in September, cost $10 and $50. That is 20% off at both ends.
A token is a chunk of a word, the unit these models read and write in. A million of them is, roughly, several hundred pages of text.
Introducing Claude Fable 5.2 and Claude Mythos 5.2, designed to tackle the most daunting technical challenges.
— Claude (@claudeai) November 9, 2026
They’re capable collaborators on ambitious projects in many fields, and show improvements in areas like coding, reasoning, and vision. pic.twitter.com/jP0AutLxXb
What’s new in Claude Fable 5.2?
Anthropic is pitching it mainly at coding and agents, the systems that work on their own for hours or days. The company says Fable 5.2 can coordinate hundreds, even more than a thousand, subagents on a single job. In a test formalising the proof of Fermat’s Last Theorem, a thousand agents finished in 30 hours. Anthropic’s first attempt, with an unreleased research model, took 11 days.
The biggest jump is in robotics. On RoboDojo, a simulated robot-control benchmark, Fable 5.2 scores 64.9% against its predecessor’s 27.6%. By our sums, that is 2.35 times higher.
It is also faster (by up to 30%, Anthropic says) and refuses fewer harmless requests: unnecessary blocks fell by about 55% in early testing.
Who can use Claude Fable 5.2?
Subscribers on the Pro, Max, Ultra, Team and Enterprise plans, and developers through the API on Anthropic’s own platform, Amazon Web Services, Google Cloud and Microsoft Foundry. A fast mode, up to 2.5 times quicker, costs double.
Mythos 5.2 is a different matter. It is the same underlying model with looser safeguards on cybersecurity and biology, and it is limited to vetted organisations.
What about safety?
Two findings in the system card stand out. In a sandbox-escape test, Mythos 5.2 used credentials or files from outside its sandbox in 3.1% of scenarios, down from 21% for Mythos 5.1. But it was also the first Anthropic model to get past a monitor reading its reasoning when told to underperform on purpose, in about one attempt in four. Anthropic says the risk of serious misalignment remains low.
The launch lands weeks after the company filed to go public, with its Wall Street debut expected soon.
By Oliver Grant
Image: Anthropic (official share image)