Community Radar.

“Claude Fable 5 feels less like a model launch and more like a preview of AI inequality”

Anthropic just released Claude Fable 5, and I think the real story is not “new model better at coding.” The real story is that frontier AI is turning into a gated utility. Public users get Fable 5, but with heavy safety routing. If the system thinks your request touches cyber, bio, chemistry, or distillation, it can kick you down to Opus 4.8. Meanwhile, selected partners get Mythos 5, which is basically the same underlying model with some safeguards lifted. So the public gets the “safe” version. Trusted institutions get the dangerous/useful version. I understand why they’re doing it. Nobody wants open access to a model that can materially improve cyberattack capability. Fine. But let’s stop pretending this is just a normal product release. This is the beginning of a two-tier AI world: one model for regular users, another model for governments, big companies, approved labs, and people inside the trust circle. Also, the subscription thing is messy. Fable 5 is included for paid plans only until June 22, then it moves to usage credits unless they have enough capacity. That tells me the economics are still ugly. These companies want everyone dependent on agents, but the best agents may be too expensive to give to normal users at normal subscription prices. My unpopular take: the next AI monopoly won’t just be about who has the smartest model. It’ll be about who gets access to the uncapped version. Everyone else gets the child-safe demo.