![Claude Fable 5]()
Anthropic has announced Claude Fable 5, the first publicly available model in its new Mythos-class AI family. The release marks a significant milestone for the company, making capabilities previously restricted to a small group of trusted organizations available to a broader audience.
From Mythos to Fable
Earlier this year, Anthropic introduced Claude Mythos Preview through its Project Glasswing initiative, providing access to a limited set of organizations, government agencies, and cybersecurity partners. The model attracted attention after helping identify thousands of high-severity software vulnerabilities, leading Anthropic to restrict wider access because of potential cybersecurity risks.
With the new launch, Anthropic is making a safeguarded version of that technology available under the name Claude Fable 5. According to the company, Fable 5 uses the same underlying Mythos-class architecture but includes additional guardrails to prevent misuse in sensitive domains such as cybersecurity, biology, and chemistry.
What Makes Claude Fable 5 Different?
Anthropic describes Fable 5 as its most capable public model to date, delivering state-of-the-art performance in:
The company says the model is more efficient than previous generations, often requiring fewer tokens to complete tasks despite higher per-token pricing.
Built-In Safety Controls
A key aspect of the release is Anthropic’s safety-first approach.
If users attempt to use Fable 5 for restricted activities such as vulnerability discovery or other high-risk cybersecurity operations, the model can automatically defer responses to a less capable model, Claude Opus 4.8, rather than executing the request. Anthropic says extensive red-teaming and adversarial testing were conducted before the public launch.
Mythos 5 Remains Restricted
Alongside Fable 5, Anthropic confirmed that Claude Mythos 5 remains available only through a controlled access program. The company plans to expand availability gradually through a trusted-access framework while continuing its Project Glasswing initiative focused on securing critical software infrastructure.
Anthropic previously reported that organizations participating in Glasswing discovered more than 10,000 high- and critical-severity vulnerabilities using Mythos-class models.
Availability
Claude Fable 5 is available immediately to Anthropic Pro, Max, Team, and Enterprise customers, as well as through major cloud platforms. Anthropic has indicated that broader access to Mythos-class capabilities will continue to expand over time through controlled programs and partnerships.