What Are Claude Mythos and Claude Fable? Anthropic's Next-Generation Models

A New Tier Above Opus

Beyond its established Opus, Sonnet, and Haiku model tiers, Anthropic has introduced a new top tier of models referred to as Mythos-class, representing what the company describes as higher intelligence than its Opus-tier models. The first model in this category, Claude Mythos Preview, was introduced as a research preview in early 2026, with access initially restricted to a small number of trusted organizations participating in what Anthropic calls Project Glasswing.

Why Access Has Been Restricted

Unlike Anthropic's mainstream model releases, which typically become broadly available shortly after launch, Mythos-class models have been held back from general release specifically because of cybersecurity concerns tied to their elevated capability level. Anthropic has indicated that a more intelligent model of this kind requires correspondingly stronger safeguards before it can be made available to all customers, and the company has described making swift progress on developing those safeguards as a precondition for wider release.

Claude Mythos 5 and Claude Fable 5

The current generation of Mythos-tier models includes two named variants, Claude Mythos 5 and Claude Fable 5, which share the same underlying model architecture. The key difference between them lies in their safety configuration: Claude Fable 5 includes additional, more restrictive safety measures specifically targeting biology, cybersecurity, and AI research and development risks, making it the more conservative of the two variants for use cases involving those especially sensitive domains.

A Temporary Suspension

Access to both Claude Mythos 5 and Claude Fable 5 was temporarily suspended in response to an export control directive, a development Anthropic addressed in a public announcement. Because export control rules and their interpretation can shift, and because Anthropic's own public statements on this topic represent the most current and authoritative source, anyone seeking the latest status of Mythos-class model availability should check Anthropic's official announcements directly rather than relying on any single fixed account of the situation.

How This Fits Into Anthropic's Broader Strategy

The existence of a restricted, higher-capability tier above Anthropic's mainstream Opus models reflects a recurring pattern in how the company has approached releasing increasingly powerful AI systems: rather than making every advance in raw capability immediately available to the widest possible audience, Anthropic has at times chosen to gate access to its most powerful systems behind additional safeguards, vetting, or specific use-case restrictions, particularly when a model's heightened capability also raises the stakes of potential misuse.

What to Watch For

As safeguards mature and any relevant export control issues are resolved, Anthropic has signaled an intention to eventually bring Mythos-class capability to a broader customer base, narrowing the gap between its publicly available Opus-tier models and its most advanced internal research. Given how quickly this particular area of Anthropic's product lineup has been evolving, readers interested in the most current details on availability, pricing, or access requirements should consult Anthropic's own product announcements rather than treating any single snapshot, including this one, as necessarily current.