Anthropic releases Fable 5 model based on Mythos

Anthropic has introduced Claude Fable 5, the first public release of its new AI model built on the Mythos architecture. The company states the new model surpasses previous versions in programming, document processing, and image analysis. However, it also comes with added restrictions in fields like cybersecurity, biology, and chemistry.

Fable 5 is highlighted for its top performance in programming, complex analyses, and handling large datasets. Despite these advancements, the model cannot process certain queries related to cybersecurity, biology, and chemistry, nor can it handle model distillation.

Anthropic initially introduced Mythos in April, with limited access granted to a select few partners due to concerns about its potential misuse in cyberattacks. By June, the company expanded testing to 15 countries and hundreds of organizations, offering a simpler version to a broader audience. Claude Fable 5 can be accessed via Anthropic API and Enterprise plans. Users with Pro, Max, Team, and Enterprise plans will have complimentary access until June 22. After this date, access will transition to a credit-based system.