Claude Fable 5 vs Mythos 5
Anthropic launched two models on June 9, 2026: Claude Fable 5 and Claude Mythos 5. They share one brain. The difference is who can use it, and what it is allowed to do.
Same model, different doors
Fable 5 and Mythos 5 are not two different models in the way Opus and Haiku are. They are the same Mythos-class system behind two different access policies. Fable 5 is the safeguarded, publicly available door. Mythos 5 is the restricted door, where some safeguards are deliberately removed for partners who need the unrestricted capability for defensive work.
Anthropic frames this as a way to release frontier capability to the public responsibly while still giving a narrow set of trusted users the full, unsafeguarded model. The split is about control, not about one model being smarter than the other.
The public, safeguarded model
Fable 5 is what you get through the API and through tools like GitHub Copilot. Its classifiers route a narrow band of requests — mainly offensive cybersecurity and biology or chemistry — to Claude Opus 4.8 instead of answering directly. Anthropic reports this fallback affects fewer than 5% of sessions, and that more than 1,000 hours of external red-teaming surfaced no universal jailbreaks.
For ordinary software, knowledge-work, analysis, and operations use, those safeguards are effectively invisible. They exist to keep the highest-risk categories out of general reach, not to limit everyday work.
The restricted, unsafeguarded model
Mythos 5 is the same model with some safeguards lifted, and it is not generally available. Anthropic initially deployed it to a small group of cyber-defenders and infrastructure providers through a program it calls Project Glasswing, in collaboration with the US government, with select biology researchers to follow through a trusted-access program.
Which one is relevant to you
For virtually every commercial use — building products, automating operations, analysis, customer-facing systems — the answer is Fable 5, because it is the one you can actually access and because its safeguards are irrelevant to that work. Mythos 5 only matters to organisations doing the specific defensive cyber or life-sciences research it was opened up for.
The more useful decision is not Fable versus Mythos, but how you deploy whichever model you use: with the right data boundaries, retention rules, and bounded authority. That is where business outcomes and business risk are actually decided.
Is Claude Mythos 5 better than Claude Fable 5?
They are the same underlying model, so capability is essentially the same. Mythos 5 has some safeguards lifted for vetted partners; Fable 5 keeps those safeguards for public use. Neither is “smarter” than the other.
Can I get access to Claude Mythos 5?
Generally no. Mythos 5 is restricted to a small set of vetted partners — initially cyber-defenders through a US-government collaboration called Project Glasswing, and later select biology researchers. The public model is Claude Fable 5.
What safeguards does Claude Fable 5 have that Mythos 5 does not?
Fable 5 routes a narrow set of sensitive requests — mainly offensive cybersecurity and biology or chemistry — to Claude Opus 4.8 instead of answering. Mythos 5 has some of those safeguards lifted for its restricted partners.
Which model should my business use?
For almost all commercial work, Claude Fable 5 — it is the publicly available version and its safeguards do not affect ordinary business tasks. Mythos 5 is only relevant to specific defensive cyber or life-sciences research.
What is Project Glasswing?
It is the program through which Anthropic initially deployed Mythos 5 to a small group of cyber-defenders and infrastructure providers, in collaboration with the US government.
We don't advise on AI. We run it for you.
Proven on your data before you commit.