Anthropic released Claude Fable 5, its first Mythos-class model. The company said new safeguards made it possible to release a model it previously deemed too risky.
Anthropic announced Claude Fable 5, its most powerful model to date. The company claims exceptional performance in software engineering, knowledge work, and vision, with the lead growing as tasks become longer and more complex.
Anthropic had previously stated that its Mythos-class models were too dangerous to release publicly due to their advanced cybersecurity capabilities. The release was made possible by new safeguards that block responses in specific high-risk areas, falling back to Claude Opus 4.8 when needed. The company reported that 95% of sessions used only Fable responses without fallback.
This release demonstrates progress in AI safety research, enabling the deployment of a model previously deemed too risky. However, the blocking of responses in sensitive domains like cybersecurity and biology may raise concerns. It represents a key case study in balancing AI capability with safety.