Claude Fable 5 is Anthropic's most capable generally available model, launched June 9, 2026 as the first of the Claude 5 family. It sits in a new Mythos class above Opus, costs $10 per million input tokens and $50 per million output, and shares its underlying model with the restricted Claude Mythos 5: the only difference between the two is safeguards.
What is Claude Fable 5?
Fable 5 opens Anthropic's Mythos class, a tier above the Opus models that led the Claude lineup through 2025 and early 2026. Anthropic positions it as state of the art on nearly all the benchmarks it tested, with the largest gains in software engineering, knowledge work, vision, and scientific research, and with longer autonomous operation than any previous Claude model.
It is not a successor to a specific Opus release. Opus 4.8 continues to exist alongside it, and Fable 5 even uses Opus 4.8 as a fallback: when one of its safety classifiers flags a session, the conversation is handed to Opus 4.8 rather than refused outright.
What is the difference between Claude Fable 5 and Mythos 5?
They are the same underlying model. The difference is who can use it and with which limits:
- Fable 5 is generally available and ships with classifier safeguards covering offensive cybersecurity, dangerous biology and chemistry, and attempts to distill the model into a competitor. Anthropic says more than 95 percent of sessions never touch the cyber fallback.
- Mythos 5 has those safeguards selectively lifted, and only for approved organizations: cybersecurity partners in Anthropic's Project Glasswing at launch, with a trusted-access program for biology researchers planned.
Anthropic's own footnote captures it: Fable comes from the Latin fabula, akin to the Greek mythos, and "the safeguards are what distinguish the two models."
How does Claude Fable 5 perform?
The launch numbers are vendor-reported, so read them as a strong opening claim rather than a settled verdict. That said, the third-party statements at launch were unusually concrete. Cognition reported Fable 5 as the highest-scoring frontier model on its FrontierCode evaluation at medium effort. Cursor called it the state of the art on CursorBench and noted it opened up long-horizon problems earlier models could not touch. Stripe described a 50-million-line Ruby codebase migration that a team had scoped at two months, completed in a day.
On the vision side, Anthropic demonstrated the model rebuilding a web app's source code from screenshots alone, and completing Pokémon FireRed through a vision-only interface.
Independent, slower-moving data takes weeks to accumulate. Community rankings for the Claude family are tracked on our Claude benchmarks page, which updates with weekly LMArena snapshots; that is the place to watch Fable 5's standing settle against the launch claims.
What does Claude Fable 5 cost and where can you use it?
API pricing is $10 per million input tokens and $50 per million output tokens, which Anthropic notes is less than half the price of the earlier Claude Mythos Preview. The API model id is claude-fable-5, available globally since launch day.
For subscribers, the rollout came in two phases: Fable 5 usage was included in subscription plans from June 9 to June 22, 2026, and from June 23 it draws on usage credits. That pricing shift matters when reading community sentiment, because a model that feels free behaves differently in the mind of a reviewer than one that meters every session.
How is the community reacting to Fable 5?
Launch-week impressions of a new frontier model are historically unreliable: the honeymoon effect inflates early ratings, and pricing changes color everything that follows. That cycle, and how to see past it, is the subject of our guide to whether Claude is getting dumber.
This site measures the reaction directly. The live Claude tracker shows how users rate their sessions day by day on a five-point scale, with optional model-version tags and short reports attached to votes; the daily record since the June 9 launch, including the June 23 switch to usage credits, is preserved in the reviews archive. How votes are collected and the sample-size caveats are documented on the about page. Other Claude versions are covered in our Claude models section.
FAQ
Is Claude Fable 5 better than Opus 4.8?
On Anthropic's reported benchmarks, yes: Fable 5 is the more capable tier. Opus 4.8 remains relevant as the fallback model Fable 5 routes flagged sessions to, and as the alignment baseline Anthropic compared against, reporting "low" misaligned behavior for the Mythos-class models, similar to Opus 4.8.
Why does Fable 5 sometimes hand my session to Opus 4.8?
Fable 5 runs classifiers for offensive cyber work, dangerous biology and chemistry, and distillation attempts. A flagged session falls back to Opus 4.8 instead of being refused. Anthropic reports that over 95 percent of sessions involve no cyber fallback, though biology and chemistry requests fell back more often in the early rollout.
Does Fable 5 replace the Opus models?
No. It starts a new tier above them. The Claude 5 family and the Opus line coexist, and Anthropic's naming makes the structure explicit: Mythos class is the ceiling, Fable 5 is its generally available form.