Anthropic launches Claude Fable 5 with new Opus safeguards

Anthropic announced on Tuesday the launch of Claude Fable 5, a Mythos-class model designed for safe general use.

The AI company claims that Fable 5’s capabilities surpass those of any model it has previously made publicly available. Fable 5 has shown “exceptional performance” in areas such as software engineering, knowledge work, computer vision, and scientific research.

It outperforms Opus models on longer and more complex tasks and can operate autonomously for longer durations than any previous Claude model.

To prevent misuse, especially in areas like cybersecurity, Fable 5 is being released with conservative safeguards. Questions on certain topics will be directed to Opus 4.8, and these safeguards are expected to activate in fewer than 5 percent of sessions on average. Most queries related to cybersecurity, chemistry, and biology will receive responses from Opus 4.8 rather than Fable 5.

Additionally, Anthropic is releasing Claude Mythos 5 to a select group of cyber defenders and infrastructure providers. While it uses the same underlying model as Fable 5, some safeguards have been lifted.

Mythos 5 is part of Project Glasswing and serves as an upgrade to the Claude Mythos Preview. Anthropic asserts that Mythos 5 possesses the strongest cybersecurity capabilities of any model available, with plans to expand its access through a broader trusted access program.

Fable 5 and Mythos 5 are priced at $10 per million input tokens and $50 per million output tokens, which is less than half the cost of the Claude Mythos Preview. Mythos 5 is exclusively available to those with access to the Mythos Preview, including Apple, one of Anthropic’s partners in Project Glasswing.

Fable 5 will be included in Pro, Max, Team, and seat-based Enterprise plans until June 22. After that date, users will need to use credits to access the model. Anthropic plans to reintroduce Fable 5 into subscription plans once capacity is sufficient.