Anthropic’s most powerful AI comes in two flavors with two different names: Claude Fable and Claude Mythos. People assume they’re two separate models. They’re not. Anthropic says plainly that “Claude Fable and Claude Mythos are the same underlying model.”
The difference is who’s allowed to use which one, and what each is allowed to help with.
Fable
Fable is the version anyone can use, through paid Claude plans and Anthropic’s developer tools. It has extra safety filters around cybersecurity and biology, the two areas where a really capable model could do serious damage in the wrong hands. When one of those filters trips, Fable hands the question off to an older, less capable Claude model instead of answering it itself.
The newest version, Fable 5.2, came out November 9. Anthropic says it’s up to 30% faster than 5.1, trips its safety filters about 55% less often on harmless requests, and costs 20% less for developers ($8 per million input tokens and $40 per million output tokens, down from $10 and $50).
Mythos
Mythos is the same model without some of those cyber and biology filters. That makes it more useful for people defending computer systems or doing lab research, and more dangerous if it got into the wrong hands. So Anthropic only offers it to vetted organizations, through verification programs for security professionals and life sciences researchers, plus its enterprise security product. Members of those verification programs also get a faster mode.
Mythos started out in Anthropic’s Project Glasswing, a program with big partners like AWS, Apple, Microsoft and CrowdStrike. In June 2026, the U.S. Commerce Department ordered both models switched off over export control concerns. Anthropic restored access in early July after those controls were lifted.
The tell
Ask whether you could sign up for it today. If yes, it’s Fable. If you’d need your company vetted first, it’s Mythos.
It’s a bit like a hardware store and a contractor’s supply counter. Same tools, same brands. But a few things are only sold behind the counter, to people who can show a license, because in the wrong hands they do real harm.
For almost everybody reading this, Fable is the one you’d actually use. Mythos mostly matters because it’s a preview of how AI companies plan to handle models that are too capable to hand out freely. That idea is spreading, and I dunno that Anthropic will be the last to do it.

