Anthropic Launches Claude Fable 5: First Public Mythos-Level AI Model

Anthropic today launched a new AI model, Claude Fable 5, marking the company's first publicly available "Mythos" tier model. According to official statements, Fable 5 surpasses all previously released models in comprehensive capability, delivering "exceptional performance" across software engineering, knowledge work, visual understanding, and scientific research. Anthropic emphasizes that the model has passed safety evaluations and is safe for general use. However, most queries involving sensitive fields such as cybersecurity, chemistry, and biology will be answered by the Opus 4.8 model rather than Fable 5, reflecting the company's deliberate balance between capability and safety.

In parallel, Anthropic has deployed the Mythos 5 model through "Project Glasswing," serving as an upgrade to the earlier Claude Mythos preview. Mythos 5 is described as the world's most capable AI model for cybersecurity, with access gradually expanding via a broader trust-based access program. Users who already had access to the Mythos preview — including Apple — can now use Mythos 5 directly. Notably, Anthropic also released another model, Claude Opus 4.8, on the same day, further enriching its product lineup.

This release marks a key step in Anthropic's high-end AI model strategy. Fable 5 occupies a position between general utility and safety control, while Mythos 5 focuses on specialized security domains. As AI model capabilities continue to rise, striking a balance between openness and risk management is becoming a central industry concern. Anthropic's tiered approach may serve as a reference for future large model product design.