Anthropic has launched its Claude Fable 5 artificial intelligence (AI) model, which it says is in the same class as Claude Mythos 5, which was also launched this morning. However, Fable 5 has a number of security features that make it safe for general use.
According to Anthropic, Claude Fable 5 outperforms all previous Claude models in almost all benchmark tests in software engineering tasks, knowledge work, vision, science, and long-term complex tasks. Meanwhile, Claude Mythos 5 is the same version but with security restrictions relaxed for select groups such as cybersecurity professionals through Project Glasswing.
In tests by some early adopters, Fable 5 can complete tasks that would have taken months in a matter of days. It is also the latest in vision with the ability to build web applications based on screenshots and extract information from graphs.
The security feature that Anthropic has placed on Fable 5 is that it switches to the Opus 4.8 model for sensitive topics such as cybersecurity, biology, chemistry, and distillation experiments. The Mythos class model has a high ability to find vulnerabilities that are then exploited. Therefore, Fable 5 is completely blocked from providing information for the purpose of carrying out cyber attacks.
Claude Fable 5 is accessible to paying customers starting today.
