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

Anthropic has officially launched Claude Fable 5, its first publicly available Mythos-level model. According to the company, Fable 5 has been classified as "safe for general-purpose use." The model delivers "exceptional performance" across software engineering, knowledge work, visual tasks, and scientific research, surpassing all previous publicly released Anthropic models. Notably, most queries related to cybersecurity, chemistry, and biology are handled by Opus 4.8 instead of Fable 5, reflecting the company’s cautious approach to sensitive domains.

Alongside Fable 5, Anthropic also released Claude Opus 4.8 and deployed Mythos 5 through the Project Glasswing initiative as an upgrade to the Claude Mythos Preview. Mythos 5 is described as "the world's most capable model for cybersecurity," with access being gradually expanded through a broader "trusted access program." Existing Mythos Preview users, including Apple, can now obtain access to Mythos 5.

This release marks a critical step in Anthropic’s commercialization of high-end AI models. The general safety assessment of Fable 5 and the specialized reinforcement of Mythos 5 highlight the company’s effort to balance openness with risk management. With partners like Apple gaining access, the model series is expected to deliver greater value across both consumer and enterprise applications.