What Is Claude Fable 5? Anthropic's Mythos-Class Model Explained
In June 2026, Anthropic released the first model from a tier above its familiar Opus line, and gave it a name that confused almost everyone: Claude Fable 5, built from a model called Mythos. Nine days later it was pulled under a federal export-control order, then reinstated three weeks after that. Here's what Fable 5 and Mythos 5 actually are, why the suspension happened, and what it costs to use one.
Fable 5 is the guardrailed public face of Mythos, a model tier Anthropic says exceeds anything it has previously released for general use.
Fable 5 Is the Public Version of a Model Called Mythos
Claude Fable 5 is Anthropic's first Mythos-class model made available to the public — a tier the company says sits above its existing Opus line and exceeds the capability of anything it has previously released generally. It launched on June 9, 2026, alongside a sibling model, Claude Mythos 5, which shares the same underlying weights but skips the public safety layer entirely.
The short version: if you're a developer or Claude subscriber, the model you can actually call is Fable 5 (API model ID claude-fable-5). Mythos 5 is not something you can sign up for — it's restricted to a small group of infrastructure providers and vetted cybersecurity researchers working with Anthropic through a program called Project Glasswing.
Same Model, Two Very Different Doors In
Fable 5 and Mythos 5 aren't two separate models trained differently — Anthropic has described them as the same underlying system with different guardrails attached. The distinction that actually matters is who can reach which version, and what happens when a request touches a sensitive area.
| Claude Fable 5 | Claude Mythos 5 | |
|---|---|---|
| Availability | Public — Claude API, Claude.ai, Claude Code, Claude Cowork, Amazon Bedrock | Limited — Project Glasswing partners only |
| Safety classifiers | Yes — can decline or reroute a request | No — safeguards removed for defensive research use |
| High-risk queries (cyber, bio, chem) | Falls back to Claude Opus 4.8 | Answers directly, under program supervision |
| Typical user | Developers, subscribers, enterprises | Infrastructure providers, cybersecurity researchers |
Anthropic has said the classifiers on Fable 5 were deliberately tuned on the conservative side to make the fast, broad release possible — meaning they'll occasionally catch a harmless request too, in exchange for reliably catching the risky ones.
The Nine Days That Got Fable 5 Pulled
A single reported jailbreak technique — not a unique Mythos-level capability — was enough to trigger a temporary federal export-control order.
Mythos disclosed publicly
Anthropic confirmed the existence of Mythos after earlier leaked documents, and launched Project Glasswing — a consortium using the model defensively to find software vulnerabilities.
Fable 5 and Mythos 5 launch
Anthropic releases Fable 5 as the first publicly available Mythos-class model, with safety classifiers attached, plus an upgraded Mythos 5 for existing Glasswing partners.
Export controls issued
The U.S. Department of Commerce restricts access after a report that Amazon researchers had prompted Fable 5 into identifying and demonstrating a software exploit.
Access restored
Commerce lifts the controls; Anthropic brings Fable 5 back online globally across the Claude Platform, Claude.ai, Claude Code, and Claude Cowork the next day.
Anthropic later noted that when it re-tested the specific exploit demonstration behind the suspension, every other model it checked — including Haiku 4.5, Sonnet 4.6, Opus 4.6 through 4.8, GPT-5.4, GPT-5.5, and Kimi K2.7 — could produce a similar result, which is why the company maintains the incident didn't reflect a uniquely Mythos-level risk.
What Fable 5 Is Actually Good At
- Long-horizon, multi-step agentic coding and research tasks
- The longer and more complex the task, the bigger its lead over Opus 4.8
- Strong early scores on software engineering and reasoning benchmarks
- Available through the same interfaces you already use — no separate signup
- Roughly double the per-token cost of Opus 4.8
- Safety classifiers occasionally decline harmless requests near sensitive topics
- Access on subscription plans has shifted between included and credit-based since launch
- Mythos 5 itself is not something individual developers can request
Frequently Asked Questions
What is Claude Fable 5?
Claude Fable 5 is Anthropic's first publicly available Mythos-class AI model, released June 9, 2026. It sits above the Opus line in capability and is aimed at long, complex reasoning and agentic coding tasks, but ships with safety classifiers that reroute high-risk queries to Claude Opus 4.8 instead.
What's the difference between Claude Fable 5 and Claude Mythos 5?
They're the same underlying model. Fable 5 is the public version with safety classifiers that can decline or reroute risky requests. Mythos 5 has those safeguards removed and is only available to vetted organizations through Project Glasswing.
Why was Claude Fable 5 temporarily suspended?
On June 12, 2026, the U.S. Department of Commerce issued export controls after a report that researchers had prompted Fable 5 into demonstrating a software exploit. The controls were lifted June 30, and access was restored globally on July 1, 2026.
More AI Model Guides, Written Clearly
Browse our full library of AI explainers, updated as models and pricing change.
More from the creator
Explore two other projects from the same creator.