What is Anthropic’s Fable 5 model?

What Anthropic launched

Anthropic released a public version of its Mythos-class model called Claude Fable 5. The company previously said its Mythos research would be too dangerous to release, but it has now moved to a more controlled public offering.

How it’s being positioned

Anthropic describes Fable 5 as a version that “comes with risks,” even though it includes guardrails. Coverage indicates that it’s designed to be usable by the public while limiting certain high-risk outcomes.

The model is offered on multiple plan tiers, and Anthropic sets a time window for public availability: Fable 5 is available on Pro, Max, Team, and seat-based Enterprise plans through June 22. After that point, using Fable 5 requires usage credits.

What safeguards are included

Anthropic has described safety mechanisms that constrain effectiveness for frontier or dangerous uses. Reported details include “invisible safeguards” using techniques such as prompt modification, steering vectors, or PEFT-style approaches to limit how the model operates.

Why it matters

Fable 5 represents a major shift in how a top-tier lab treats “frontier-capability” access: rather than keeping a powerful model entirely behind closed doors, Anthropic is moving to broader availability while trying to manage misuse risk.

It also signals competitive momentum in the AI model landscape—by letting more users test a powerful Mythos-derived system, Anthropic increases real-world demand and adoption while still aiming to reduce catastrophic abuse pathways.

What to watch next

Because the company explicitly acknowledges risks and describes fallback behavior, the most important operational question for users is how often the safeguards trigger and what the model refuses or changes under high-risk prompts.