Anthropic Rolls Out Claude Mythos 5 AI Model—Along With the Safer Fable 5 for the Public

Anthropic on Tuesday announced the official launch of Claude Mythos 5, a restricted-access AI model the company says has unprecedented cybersecurity capabilities, alongside Claude Fable 5, a version designed for broader public use.

In a blog post, Anthropic said Mythos 5 will initially be available only to approved cybersecurity organizations, critical infrastructure operators, government partners, and selected life sciences researchers. The company says the Mythos-class AI models “have reached a threshold where they present significant risks.”

“In April, we began Project Glasswing, releasing the first Mythos-class model (Claude Mythos Preview) to only a limited group of cyber defenders and critical software infrastructure providers,” Anthropic wrote. “When we did so, we stated that we hoped to eventually release Mythos-level capabilities to all our users, so long as we had developed new safeguards that were strong enough to reliably prevent misuse.”

The launch follows months of concern over Mythos’ ability to identify software vulnerabilities and conduct complex cybersecurity operations after news of the AI leaked online in March. Earlier this year, the U.K.’s AI Security Institute reported that a preview version became the first AI model to complete a 32-step corporate network intrusion exercise without human assistance. In April, Mozilla reported that a preview version of Mythos discovered over 271 vulnerabilities in the Firefox browser.

Until now, Anthropic has restricted access to Mythos through its Project Glasswing program while researchers evaluated the risks. In a follow-up post on X, Anthropic said it would continue to expand access to Mythos “through a broader trusted access program,” for cybersecurity work and biomedical research.

The launch of Mythos also comes as tensions between Anthropic and President Donald Trump’s administration appear to be easing after a public feud over the use of Claude for military operations.

“In consultation with the U.S. government, we plan to steadily expand access to Claude Mythos 5, continuing our periodic addition of new partners, as well as pursuing a trusted access program that allows cybersecurity organizations to apply in a more systematic manner,” Anthropic wrote, adding that users of Mythos Preview can upgrade to Mythos 5.

Earlier this month, press reports said that Anthropic had embedded engineers at the National Security Agency to help deploy its Mythos AI model for cybersecurity operations, reportedly targeting China and Iran.

Alongside Mythos 5, Anthropic released Claude Fable 5, a Mythos-class AI model that uses the same core design of Mythos 5 but with additional safeguards for the general public.

“Fable 5’s capabilities exceed those of any model we’ve ever made generally available,” Anthropic wrote. “It is state-of-the-art on nearly all tested benchmarks of AI capability, showing exceptional performance in software engineering, knowledge work, vision, scientific research, and many other areas.”

According to Anthropic, Mythos 5 and Fable 5 achieved top scores on several software engineering, cybersecurity, biology, and reasoning benchmarks. The company reported a 78% score on ExploitBench, which measures vulnerability discovery and exploitation, and an 88% score on the Terminal-Bench coding benchmark. Anthropic said the results show the models can handle longer and more complex tasks than previous Claude systems.

While Anthropic said Fable 5 performs especially well on lengthy, complex tasks, in order to prevent misuse, requests involving cybersecurity, biology, chemistry, or AI model replication are handled by Claude Opus 4.8—a safeguard the company said is triggered in fewer than 5% of sessions.

“Releasing a model this capable comes with risks,” Anthropic wrote. “Without safeguards, Fable 5’s capabilities in areas like cybersecurity could be misused to cause serious damage.”

The launch comes as AI developers face growing questions about how to release increasingly capable systems. Anthropic has argued that Mythos could help defenders find and fix vulnerabilities before attackers exploit them. At the same time, critics, including OpenAI CEO Sam Altman, have accused the company of using “fear-based marketing” to sell its AI models.