Anthropic Releases Claude Fable 5 and Mythos 5: One Model, Two Access Tiers

Anthropic has released Claude Fable 5 to general availability and simultaneously announced Claude Mythos 5 — a version of the same base model with a fundamentally different security profile, available exclusively through the closed Project Glasswing program.

Two products, one core

The key feature of this release — Fable 5 and Mythos 5 are built on the same base weights. The difference is not in architecture or training, but in the set of classifiers and security policies applied on top.

Claude Fable 5 enters general availability in standard configuration: with a full stack of protective filters optimized for a broad audience. This is the version for API developers, enterprise clients, and end users, undergoing Anthropic's standard security assessment process.

Claude Mythos 5 is fundamentally different. The same model, but with intentionally removed restrictions in the cybersecurity domain. It does not enter general sale.

Project Glasswing: the closed channel

Project Glasswing is Anthropic's program for trusted security partners. Through it, the company provides select organizations with access to the model with expanded capabilities — in areas where standard restrictions would create obstacles for professional work.

Typical program participants:

• Offensive security researchers testing real attack vectors
• Red team specialists working with system vulnerabilities
• Government agencies in cybersecurity
• Security professionals at major infrastructure companies
• Academic groups studying AI threats

Access to the program is tightly controlled: Anthropic vets partners before providing access to Mythos 5.

A new tier in the model hierarchy

The launch of Mythos 5 means that Anthropic is formalizing a new level in its product line — Mythos-class. This is not a marketing name, but a separate deployment mode with different access conditions, specialized documentation, and usage monitoring on Anthropic's side.

Previously the hierarchy was clear: Haiku, Sonnet, Opus. Fable 5 took the position of the top public model. Mythos-class introduces a fundamentally new dimension — not by power, but by usage policy.

"Same model, different guarantees" — this is a statement about the

approach: security as configuration on top of base weights, not as an inherent property of the model itself.

Response to market security pressure

Anthropic has long faced criticism: its models are too cautious where security specialists want to work with real threats. Competing solutions — open models or less restrictive providers — capture this segment simply because they impose fewer restrictions.

Project Glasswing and Mythos 5 are a direct response to this challenge. Anthropic is not lowering the bar for all users, but creating a controlled channel for professionals who need capabilities beyond standard profiles.

What this means

Anthropic is betting on differentiation: the broad market gets Fable 5 with full protection, trusted security partners get Mythos 5 with untethered hands. This is an attempt not to cede the security market to competitors while preserving the company's reputation as one that prioritizes safe AI. The question is how tightly the control over Glasswing participants will remain as the program grows.