Anthropic Launches Fable and Mythos 5.2 With Major Science Claims

Anthropic has launched Claude Fable 5.2 and Claude Mythos 5.2, pitching them for difficult, open-ended technical work. The company says the models improve coding, reasoning and visual tasks while generating output up to 30% faster than their 5.1 predecessors. Its announcement also makes unusually ambitious claims about work in mathematics and black-hole physics, which deserve to be read with some precision.

According to Anthropic’s 9 November release, Fable and Mythos 5.2 share the same underlying model. Fable has stronger safeguards against misuse in cybersecurity and biology, allowing general access. Mythos is intended for vetted organisations working on beneficial security and life-sciences applications. That is the same access distinction that TechBooky examined when the 5.1 models launched, now applied to a more capable version.

Anthropic lists standard API prices of $8 per million input tokens and $40 per million output tokens for Fable 5.2 and Mythos 5.2, lower than the previous versions, although Mythos access remains restricted. An optional fast mode promises token generation up to 2.5 times faster at $16 per million input tokens and $80 per million output tokens. The speed claims are Anthropic’s, and real results will vary with the task and deployment.

Fable 5.2 is available through Claude and the Claude Platform, as well as cloud partners including Amazon Web Services, Google Cloud and Microsoft Azure. Anthropic recommends Opus 5.5 as a starting point for most everyday coding and knowledge work, with Fable or Mythos reserved for more demanding projects. Security professionals and life-sciences researchers must apply through Anthropic’s verification programmes to use Mythos 5.2.

Anthropic says Fable 5.2 completed calculations addressing two open questions in theoretical black-hole physics during prerelease testing. It points to preprints for the results. A preprint makes a finding available for scrutiny, but it is not the same thing as broad independent acceptance by the research community. The company’s description of the results should therefore be reported as a research claim, not as a settled account of black-hole physics.

The Fermat’s Last Theorem result is also easy to overstate. Fable 5.2 did not discover a new proof of the centuries-old theorem. A team of roughly 1,000 model agents translated an existing proof into Lean, a formal language in which a computer can check each step. Anthropic says this took 30 hours with minimal human intervention, compared with 11 days for an earlier research version of Claude using a purpose-built workflow and more human guidance. The work is impressive as proof formalisation, but it should not be described as solving Fermat’s Last Theorem for the first time. Anthropic has published a separate account of the formalisation work.

Anthropic also says pre-release versions assisted with software migrations and hardware design. These examples show how the company wants to position the models as technical collaborators rather than general chatbots. They do not, on their own, tell customers how reliably the systems will perform on unrelated projects. Independent replication and careful evaluation remain important, particularly where research claims or critical infrastructure are involved.

The launch reinforces an increasingly important question for frontier AI companies: how much of a model’s capability should be available to everyone? Anthropic says Mythos 5.2 is especially strong at finding and reproducing software vulnerabilities, while Fable’s safeguards will block many sensitive cyber tasks. That limits some legitimate use alongside potential misuse. Its verified-access scheme for defenders is the company’s attempt to manage that trade-off.

For buyers, the near-term story is a faster, cheaper frontier model with a clear access boundary. For scientists, the more interesting test comes next: whether the formalised mathematics and physics results withstand close review, and whether comparable gains appear outside Anthropic’s own demonstrations.