Anthropic just shipped the public version of Mythos

Hook

Two days ago this site walked you through Claude Mythos, the model Anthropic built that was so good at breaking software it refused to sell it to anyone. On June 9 it sold a version of that model to everyone. The assumption that Mythos-class power stays locked behind a government program just expired.

The news

Anthropic released Claude Fable 5, which it calls a Mythos-class model made safe for general use, "state-of-the-art on nearly all tested benchmarks." It costs $10 per million input tokens and $50 per million output tokens. On a narrow set of sensitive topics (cybersecurity, biology, chemistry) the safeguards hand the query off to Claude Opus 4.8 instead, which Anthropic says happens in under 5% of sessions. It also shipped Claude Mythos 5, the same underlying model with safeguards lifted, but only to Project Glasswing partners and a few biology researchers. Anthropic's announcement is here.

On your desk

The most capable model Anthropic has ever made generally available is now sitting behind your existing Pro or Max plan, free through June 22 before usage credits kick in. Stripe says it "compressed months of engineering into days." For an operator, the takeaway is plain: the frontier you read about last week is the default you build on this week.

What you should do next

Open Claude, switch the model to Fable 5, and re-run the hardest task you handed it last month.