Anthropic’s new AI models explained: What are Fable 5 and Mythos 5, and who gets access?

Anthropic has introduced a new category of AI systems called “Mythos-class” models, marking what the company describes as a significant jump beyond its earlier Opus and Sonnet families. The launch includes two closely related models, Claude Fable 5 and Claude Mythos 5, but with very different access and safety controls attached to each. According to Anthropic, the two models share the same underlying architecture, with the primary distinction being the safeguards placed around Fable 5.

What is a Mythos-class model?

Mythos-class is Anthropic’s newest and most capable model category. The company says these models are designed for long-horizon tasks, software engineering, complex reasoning, scientific work, and knowledge-intensive workflows.

The Mythos family first surfaced publicly earlier this year through the restricted-release Claude Mythos Preview programme. Anthropic had initially declined to make the model broadly available, citing concerns about its unusually strong cybersecurity capabilities.

With this launch, Anthropic is making a Mythos-class model available to the public for the first time.

What is Claude Fable 5?

Claude Fable 5 is Anthropic’s first publicly available Mythos-class model.

The company describes it as its most capable generally accessible AI model to date. Fable 5 is aimed at software development, analytical work, research tasks and complex agentic workflows that require sustained reasoning over long periods.

Unlike previous Claude models, Fable 5 has been built with additional safety systems that monitor certain categories of potentially dangerous requests. Anthropic says the model underwent extensive internal testing, external red-teaming exercises and jailbreak evaluations before release.

What is Claude Mythos 5?

Claude Mythos 5 is effectively the less restricted version of the same underlying system.

Anthropic says Mythos 5 and Fable 5 share identical core model capabilities. However, Mythos 5 does not include some of the additional safety restrictions that have been added to Fable 5.

The model remains tightly controlled because of concerns around advanced cybersecurity applications. Earlier versions of Mythos demonstrated an ability to identify and exploit software vulnerabilities at a level that prompted Anthropic to limit distribution.

Why did Anthropic create two versions of the same model?

The split appears to be Anthropic’s attempt to balance capability and safety.

Fable 5 serves as the public-facing version, while Mythos 5 is reserved for organisations that require the model’s full capabilities. Rather than permanently restricting access to the most capable model, Anthropic is pursuing a controlled-distribution approach through trusted-access programmes and cybersecurity partnerships.

What guardrails exist in Fable 5?

Anthropic’s key safeguard is a routing system for high-risk requests. When users attempt tasks involving sensitive cybersecurity, biology or chemistry domains, Fable 5 can refuse the request or automatically defer the interaction to the less capable Claude Opus 4.8 model. Anthropic says this allows users to retain general functionality while preventing access to potentially dangerous capabilities.

The company says it also conducted adversarial testing, bug-bounty programmes and external red-team exercises before launch, claiming testers were unable to discover a universal jailbreak capable of bypassing the safeguards.

Who can access Fable 5?

Fable 5 is available immediately to paid Claude users, including Pro, Max, Team and Enterprise customers. The model is also being integrated into developer platforms and cloud ecosystems, including GitHub Copilot and cloud providers that offer Anthropic models through managed AI services.

Who can access Mythos 5?

Access to Mythos 5 remains highly restricted. Anthropic says the model will primarily be distributed through trusted-access programmes and Project Glasswing, the company’s initiative focused on securing critical software infrastructure. Earlier versions were available only to a limited group of vetted organisations, governments, security researchers and infrastructure operators.

The company has indicated that access will expand gradually but has not provided a broad public-release timeline.

How much do the models cost?

Anthropic has priced both Fable 5 and Mythos 5 at $10 per million input tokens and $50 per million output tokens. The company says this is less than half the pricing of the earlier Mythos Preview programme, although it remains roughly twice as expensive as some of its previous Opus models.

Why is cybersecurity at the centre of this launch?

Much of the discussion around Mythos stems from its performance in vulnerability discovery and software security tasks.

Anthropic previously said Mythos Preview was capable of identifying critical software flaws at a level that could reshape both cyber-defence and cyber-offence. That capability led the company to launch Project Glasswing, a programme intended to help governments and major organisations use the technology to strengthen critical infrastructure rather than expose it. The launch of Fable 5 reflects Anthropic’s belief that these capabilities can now be made available more broadly, provided additional safeguards remain in place.

What does this launch mean for Anthropic?

The introduction of Fable 5 marks the first time Anthropic has publicly released a Mythos-class model after months of arguing that the technology was too powerful for unrestricted deployment.

The new model launches also come at a time as the company prepares for one of the most closely watched public listings in technology. The Claude-maker confidentially filed for an initial public offering earlier this month, shortly after raising $65 billion in fresh funding that pushed its valuation to about $965 billion, ahead of rival OpenAI.