Claude Fable 5 Explained: What Makes Anthropic’s New AI So Powerful?

Anthropic, the company behind the Claude family of AI models, has introduced Claude Fable 5, expanding access to its most advanced Mythos-class artificial intelligence technology. Simultaneously, it has launched Claude Mythos 5, which remains restricted to vetted organisations under its Project Glasswing programme.

The launch marks a significant shift for Anthropic. Until recently, Mythos-class systems remained confined to restricted environments because of concerns over their advanced capabilities. With Fable 5, Anthropic is making its advanced AI technology available to more users while continuing to place restrictions on capabilities it considers sensitive.

The release follows the debut of Mythos Preview two months ago, a model that attracted attention in technology and cybersecurity circles for its ability to identify software vulnerabilities.

What Is Claude Fable 5?

Claude Fable 5 is Anthropic's latest public artificial intelligence model and the first Mythos-class system available to general users. Mythos-class refers to Anthropic's highest-capability AI systems, designed for advanced reasoning, coding and technical analysis.

The company describes Fable 5 as a model that can handle longer, more complex tasks that require sustained reasoning. Instead of answering a single prompt, it can work through multi-step assignments and keep context through different stages of a task.

Fable 5 is built to solve tasks that involve multiple stages, unlike previous AI models that were mostly built for single prompt responses. It can plan, carry out and refine work while maintaining context throughout a project.

What Can Fable 5 Do?

Fable 5 is aimed at users handling complex work such as coding, technical analysis and research. According to Anthropic, the model can process different kinds of information and is designed to manage tasks that require planning, reasoning and sustained context.

The company says performance improvements become most visible when tasks increase in complexity. Rather than delivering isolated answers, the model is designed to follow a chain of actions and maintain consistency as work progresses.

According to the company, Fable 5 performed strongly across several coding and reasoning benchmarks, with the biggest improvements seen in more complex tasks. The company highlighted results from technical and problem-solving evaluations where the model showed stronger performance than its earlier systems.

For users, the launch is significant because technology that was previously limited to restricted programmes is now available more broadly. While certain safeguards remain in place, Fable 5 is Anthropic's most advanced AI model released for public use so far.

Why Anthropic Created Two Versions

Rather than choosing between a full public launch and complete restriction, Anthropic has opted for a middle path. The company believes advanced AI systems can reach a broader audience without exposing every capability that may create security or safety concerns.

This approach has resulted in two closely related models. Claude Fable 5 is designed for public use and comes with safeguards that limit how it behaves in sensitive situations. Claude Mythos 5, meanwhile, remains available only to approved partners that already participate in Anthropic's controlled access programmes.

Although both models are built on the same underlying technology, Anthropic says the difference lies in how they respond to certain high-risk requests rather than in their core capabilities.

While Fable 5 is available more broadly, Anthropic continues to limit access to Mythos 5. The restricted model is intended for organisations operating in specialised environments where advanced technical capabilities can be monitored and evaluated under controlled conditions.

Fable 5 Safety Features Explained

The decision is closely linked to the origins of the Mythos programme. Earlier versions were developed with a strong focus on cybersecurity, particularly in areas involving vulnerability discovery and system analysis.

As those capabilities improved, concerns grew around how they could be used outside defensive research settings. The same tools that help security teams identify weaknesses can also provide insight into how those weaknesses might be exploited.

Fable 5 approaches those situations differently. When users enter areas that Anthropic considers sensitive, the public model applies safeguards that can restrict, limit or block certain responses altogether.

To support wider deployment, Anthropic has added multiple layers of protection designed to detect potentially risky requests and adjust the model's behaviour before detailed responses are generated.

In some cases, the system may refuse to answer specific questions. In others, it may rely on alternative models that operate with stricter safety controls, reducing the likelihood of misuse.

The AI start-up says these measures significantly lower the model's ability to complete offensive cybersecurity tasks when compared with unrestricted Mythos systems. However, some legitimate technical queries may receive reduced responses if the system classifies them as sensitive.

The launch also comes as Anthropic expands access to its AI products and prepares for its next phase of growth. According to a Reuters report, the company has confidentially filed for a US initial public offering.

Fable 5 is available through Anthropic's application programming interface (API) and enterprise plans, while Mythos 5 remains restricted to approved partners.