Anthropic has launched Claude Fable 5, a Mythos-class model designed to bring some of the company’s most advanced capabilities to broader users, while adding a new safety layer for high-risk domains.
The launch matters because Mythos had previously been treated as too sensitive for broad release. In April, Anthropic limited Mythos access to a small group of cyber defenders and critical infrastructure providers through Project Glasswing. With Fable 5, the company is attempting to make that class of capability more widely available.
What Claude Fable 5 is
Claude Fable 5 is, according to Anthropic, the most capable model it has ever made generally available. The company describes it as a Mythos-class model made safe for general use.
The improvements focus on long, complex tasks:
- Software engineering and large-scale code migrations.
- Knowledge work, document reasoning, charts, and tables.
- Vision, including scientific figures and UI reconstruction from screenshots.
- Memory and long-context performance across multi-step work.
- Scientific research workflows that require sustained strategy.
The core claim is simple: the longer and more complex the task, the larger Fable 5’s lead over previous Claude models.
What Claude Mythos 5 is
Claude Mythos 5 uses the same underlying model as Fable 5, but with safeguards lifted in some areas for trusted organizations. Anthropic is initially deploying it as an upgrade to Claude Mythos Preview through Project Glasswing, in collaboration with the US government.
Anthropic positions Mythos 5 as having the strongest cybersecurity capabilities of any model in the world. It will not be broadly available at launch; the company plans to expand access through a broader trusted access program.
The key difference: broad capability vs. trusted access
| Model | Access | Safety model |
|---|---|---|
| Claude Fable 5 | Paid users, Enterprise, and API | Conservative safeguards; high-risk requests may fall back to Claude Opus 4.8 |
| Claude Mythos 5 | Approved cyber defenders and critical infrastructure providers | Same base model, with restrictions lifted in some areas |
In short: Fable 5 is the Mythos-class model most people can use. Mythos 5 remains reserved for higher-trust contexts.
New safeguards
Anthropic says Fable 5 is powerful enough to pose serious misuse risks, especially in domains like cybersecurity, biology, chemistry, and model distillation.
If a request falls into a high-risk area, Fable 5 may block the response and route the answer through Claude Opus 4.8 instead. Anthropic says the safeguards are conservative and may create false positives, but they trigger in less than 5% of sessions on average.
There is also an enterprise policy shift: TechCrunch reported that Anthropic will require 30-day traffic retention for Fable 5 and Mythos 5, even for customers with previous zero-retention agreements. Anthropic says the data will not be used for training, but for defending against novel attacks, jailbreaks, and reducing false positives.
The benchmark chart helps put the jump in context: Fable 5 and Mythos 5 sit as Anthropic’s new reference point for complex tasks, especially when the work requires sustained reasoning, tool use, and longer-horizon autonomy.
Capabilities to watch
Software engineering
Fable 5 is aimed at longer-horizon coding tasks. In early testing, Stripe said the model compressed a codebase-wide migration that would normally take a team more than two months into a one-day process inside a 50-million-line Ruby codebase.
It also shows strong performance on coding evaluations such as FrontierCode and CursorBench, where production-quality implementation matters more than simply passing a toy benchmark.
Knowledge work
Anthropic highlights gains in finance, document reasoning, chart/table interpretation, and complex analysis. Hex said Fable 5 was the first model to score above 90% on its core benchmark for long-running analytical tasks.
Vision
Fable 5 improves on difficult visual work: extracting numbers from scientific figures, understanding screenshots, and rebuilding apps from UI images. Anthropic also showcased the model completing Pokémon FireRed with a minimal vision-only harness.
Science and biology
Mythos 5’s examples include protein design, molecular biology hypotheses, and large-scale genomics research. These use cases help explain why Anthropic is splitting the public Fable release from the more restricted Mythos access program.
Pricing and availability
Anthropic lists Claude Fable 5 and Claude Mythos 5 at:
- $10 per million input tokens
- $50 per million output tokens
That puts the models in a premium category. TechCrunch reported that Fable 5 is available through the Claude API and consumption-based Enterprise plans; subscription access for Pro, Max, Team, and seat-based Enterprise plans is rolling out in stages, with changes after June 22.
Why it matters
This release captures a tension we will see more often: frontier models are becoming useful enough to unlock new workflows, but capable enough to require tighter safety controls.
Fable 5 is important because it shows one possible path: broad access to Mythos-class capabilities, but with filters, fallback behavior, and monitoring. Mythos 5 represents the trusted-access path for use cases that need fewer restrictions and more specialized capability.
For developers, product teams, and enterprises, the takeaway is clear: the most capable models are becoming available, but price, retention policies, and safety limits are now part of the technical decision.
Sources
- Anthropic: Claude Fable 5 and Claude Mythos 5
- Claude Help Center: Release notes — Claude Fable 5 launch
- TechCrunch: Anthropic’s Claude Fable is a version of Mythos the public can access today
- CNBC: Anthropic releases Mythos-like AI model to the public