Anthropic introduced Claude Fable 5.2 and Claude Mythos 5.2 on 9 November, making Fable generally available while limiting initial access to Mythos to vetted organisations, The Verge reported. The release follows the September arrival of Fable 5.1 and Mythos 5.1. Anthropic says the new models produce output faster and perform better on coding, reasoning and vision tasks, although the two models are intended for different kinds of access.
Key points
- Fable 5.2 is generally available; Mythos 5.2 is restricted to vetted organisations.
- Anthropic says both models produce output faster than their 5.1 predecessors.
- Fable 5.2 is priced at $8 per million input tokens and $40 per million output tokens.
- The company says it tested the models for safety and developed monitoring tools for use after release.
Fable 5.2 pairs faster output with lower prices
Anthropic says Fable 5.2 and Mythos 5.2 generate output 30% faster than their predecessors and cost 20% less than the 5.1 generation, CNBC reported. The speed figure has a qualification elsewhere: 9to5Mac quoted Anthropic describing the models as producing output “up to 30% faster”. Neither wording turns the figure into a promise that every response will arrive at that speed.
Fable 5.2 costs $8 per million input tokens and $40 per million output tokens. Fable 5.1 cost $10 and $50 respectively, Yahoo Finance reported. Tokens are pieces of text counted on both sides of an exchange: the material sent to a model is input, and its reply is output. The separate prices matter because reading a long document and writing a long answer put different amounts through those two parts of the meter.
Reworking a difficult piece of software could cost less at Fable 5.2’s stated rates than at Fable 5.1’s, while a faster response could make repeated attempts less time-consuming. Those possibilities depend on how much material the task sends and how much the model writes back.
Anthropic describes Fable 5.2 as suited to difficult software engineering and says it can coordinate teams of agents on long-running projects. The company also says prerelease use of the two models included software migrations, designs for three working electronic devices and work on problems in black-hole physics. Those accounts concern work Anthropic describes from before launch, rather than a published comparison of the models across those tasks.
Mythos 5.2 remains behind verification programmes
Mythos 5.2 is available to vetted organisations through Anthropic’s Life Sciences and Cyber Verification Programs. Anthropic introduced Mythos in April and restricted access because of its advanced cybersecurity capabilities; it introduced Fable, a public version with additional safeguards, two months later. The same distinction persists in this release: an organisation can obtain Fable 5.2 generally, while access to Mythos 5.2 depends on admission to one of those programmes.
The arrangement has a practical consequence for companies building on Claude. They can consider Fable 5.2 for broadly available services, including work that Anthropic describes as technically demanding, but Mythos 5.2 is available only through the restricted route. Anthropic’s earlier cyber programme brought together 11 partners and a free scanner, another instance of the company organising access to cybersecurity work through a programme.
Dianne Penn, Anthropic’s head of product management for research, told CNBC that Fable 5.2 is intended chiefly for complex, open-ended work in engineering, mathematics and robotics. She said Opus 5.5 was better suited to daily tasks. Anthropic had released Opus 5.5 with lower prices and stronger coding scores, giving customers another model to consider alongside Fable rather than making the new release a replacement for every task.
Amodei’s September call and Anthropic’s safety tests
The 5.2 launch is Anthropic’s first release of frontier models since chief executive Dario Amodei called in September for a slower pace of AI development, according to CNBC. Penn described the company’s position to the broadcaster as “pacing, not pausing”. Anthropic says it put the models through safety and alignment testing and developed new evaluations and tools to monitor them after release.
That combination places the new models in the middle of a choice Anthropic is making in public: it is releasing more capable systems while describing additional checks around their deployment. Fable’s general availability and Mythos’s restricted access are concrete parts of that choice. The company’s account of its testing is separate from the accounts of what its engineers accomplished with the models during prerelease use.
Anthropic is preparing for an initial public offering, for which it is seeking a valuation of $1 trillion to $1.5 trillion, down from a previously discussed $2 trillion target. In the company’s amended prospectus, Anthropic said it generated $33.24 billion in revenue for the nine months ended 30 September, up roughly 1,200% from the same period a year earlier.
