Claude Fable 5: Anthropic’s Mythos-Class Model, Explained
In June 2026, Anthropic released the most capable AI model it has ever made available to the public: Claude Fable 5. It created an entirely new tier above the familiar Opus line, got suspended by the US government within days, came back with stronger guardrails, and is now — briefly — included in paid Claude plans. Here’s the whole story, and what it means in practice.
A new class above Opus
For years the Claude family had three tiers: Haiku (fast), Sonnet (balanced), Opus (most capable). Fable 5 introduced a fourth tier that Anthropic calls the Mythos class. According to Anthropic, the model is state-of-the-art on nearly all major capability benchmarks, and its advantage over other models grows as tasks get longer and more complex — exactly the kind of extended, autonomous work that AI agents do. Early testers reported striking results; Stripe, for example, said the model completed a codebase-wide migration in a day that would otherwise have taken a team over two months.
Fable vs Mythos: two names, one model
Here’s the interesting design decision. Fable 5 and its sibling Claude Mythos 5 are the same underlying model. The difference is safeguards: Fable 5 ships with the strongest safety classifiers Anthropic has ever applied — particularly around offensive cybersecurity — and is available to everyone. Mythos 5 has those restrictions lifted in some areas and is available only to a small set of approved organizations, initially cyber-defenders and infrastructure providers working with the US government under Project Glasswing. In other words: one brain, two levels of trust.
The export-control drama
Days after the June 9 launch, the US government applied export controls to both models, citing a report that researchers had found a way to bypass Fable 5’s safeguards on vulnerability-related tasks. Because the order required restricting access by foreign nationals and there was no reliable way to verify nationality in real time, Anthropic suspended both models for everyone. After a two-week review — during which Anthropic’s testing showed that several less capable models could reproduce the same results — the controls were lifted, and Fable 5 returned globally on July 1 with a new set of cybersecurity classifiers.
Specs, pricing, and the current promo
The numbers are serious: a 1-million-token context window by default, up to 128k output tokens per request, and API pricing of $10 per million input tokens and $50 per million output tokens — premium territory, though notably cheaper than the earlier Mythos Preview. For subscribers, there’s a limited-time deal: through July 12, 2026, Pro, Max, Team, and select Enterprise plans include Fable 5 for up to 50% of weekly usage limits. After that, it moves to prepaid usage credits, though Anthropic has indicated it aims to bring it back into standard subscriptions when capacity allows.
What this means for regular users
Two practical takeaways. First, if you have a paid Claude plan, this window is worth using deliberately — point Fable 5 at your hardest problems (large codebases, long documents, complex analysis) while it’s included. Second, expect occasional refusals: the safety classifiers are deliberately tuned cautiously, and some legitimate technical requests may get declined and fall back to Opus. That’s the trade-off Anthropic chose — err on the side of blocking — to make a model this capable broadly available at all. For those of us building AI into daily workflows, Fable 5 is a preview of where the whole field is heading: models that don’t just answer, but work.
