Anthropic releases Claude Fable 5 and Mythos 5 with distinct safety configurations

• •Anthropic launched Claude Fable 5 for general availability alongside a new restricted-access model, Claude Mythos 5.
• •MarkTechPost reports that while both models share the same underlying architecture, they differ significantly in their integrated safety classifiers.
• •Claude Mythos 5 is currently limited to the 'Project Glasswing' initiative, which notably removes certain cybersecurity safeguards present in the Fable version.
• •It remains unclear how Anthropic plans to manage the risks associated with the reduced safety protocols in Mythos 5 as it scales.

Anthropic has introduced two new model versions, Claude Fable 5 and Claude Mythos 5, which share a core architecture but operate under different safety frameworks. According to reporting from MarkTechPost, Fable 5 is available for general use with standard safety classifiers, while Mythos 5 is confined to the experimental Project Glasswing. This restricted version operates with fewer cybersecurity safeguards enabled, marking a shift in how the company deploys its high-capability models. Whether this tiered approach effectively balances innovation with safety will depend on the real-world performance of the Project Glasswing cohort.