Anthropic announced Tuesday the public release of Claude Fable 5, a Mythos-class AI model available to paid subscribers and enterprise customers.
The release marks the first time the company has made a Mythos-class model available to the general public. This move signals a shift in the company's approach to deploying high-capability models that were previously deemed too risky for wide distribution.
The announcement follows a limited rollout that began two months ago [1]. By expanding access, Anthropic aims to integrate its most powerful capabilities into professional and corporate workflows, while maintaining strict oversight of the model's output.
To manage the risks associated with the model's power, the company implemented new security measures. These safeguards are designed to prevent the AI from being used for malicious purposes, such as coordinating cyber-attacks.
"We've built new safeguards that block responses in specific high‑risk areas," Dario Amodei, Anthropic co-founder, said.
These restrictions are intended to balance the utility of the Mythos-class architecture with the necessity of public safety. The company believes these blocks allow for a broader release without compromising security standards.
An Anthropic spokesperson said the capabilities of this new release "exceed those of every model we've previously made generally available."
While the company has not detailed the specific technical benchmarks of Fable 5, the Mythos classification denotes a tier of AI performance above previous Claude iterations. The transition from a restricted preview to a paid subscription model allows the company to monitor usage patterns across a larger, diverse user base.
“"exceed those of every model we've previously made generally available,"”
The public debut of a Mythos-class model indicates that Anthropic believes its safety alignment techniques have reached a threshold where the utility of high-reasoning AI outweighs the potential for misuse. By restricting the release to enterprise and paid users, the company creates a controlled environment to stress-test these safeguards before any potential free-tier rollout.
