Anthropic is giving cybersecurity professionals access to its most powerful version while offering the general public a restricted alternative designed to prevent misuse.
The AI company announced the launch of Claude Mythos 5, an upgraded version of its cybersecurity-focused model available only to trusted organizations, alongside Claude Fable 5, a public-facing version that includes extensive safety controls.
Anthropic described Mythos 5 as having the "strongest cybersecurity capabilities of any model in the world." However, access remains limited through Project Glasswing, a partnership launched earlier this year with governments, infrastructure operators, and security organizations to identify and patch vulnerabilities before malicious actors can exploit them.
The company first introduced Mythos Preview in April but restricted its availability after internal testing showed unprecedented performance in vulnerability discovery and cybersecurity tasks. Rather than releasing it broadly, Anthropic partnered with a small group of vetted organizations, later expanding the program to roughly 150 organizations across more than 15 countries.
Now, those partners will receive an upgrade to Mythos 5, which Anthropic says retains access to advanced cybersecurity capabilities that have been intentionally limited in the public version. According to the company, broader access may eventually be offered through a "trusted access" program, although no timeline has been announced.
For everyone else, Anthropic is releasing Claude Fable 5, a model built on the same underlying technology as Mythos 5 but equipped with what the company calls "robust safeguards." Those safeguards are designed to prevent users from obtaining information that could facilitate cyberattacks or assist with sensitive biological and chemical research.
When users attempt to access restricted capabilities, Fable 5 automatically redirects those requests to Claude Opus 4.8, an older and less capable model. Anthropic says extensive red-team testing failed to uncover universal methods for bypassing the new protections. The company also claims the system can detect attempts to circumvent its restrictions and respond accordingly.
Despite the limitations, Fable 5 remains Anthropic's most capable publicly available model to date. The company says it excels in software engineering, analytical reasoning, scientific research, and visual tasks while maintaining restrictions around areas considered especially dangerous. Pricing starts at $10 per million input tokens and $50 per million output tokens, making it Anthropic's most expensive offering.
"Fable 5's capabilities exceed those of any model we've ever made generally available. 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. The longer and more complex the task, the larger Fable 5's lead over our other models," the company wrote.
