Anthropic Launches Claude Fable 5: Its Most Powerful Public Model Yet, With the Cyber Edges Sanded Off

Anthropic has done the thing it spent two months refusing to do: it put a Mythos-class model in the public's hands. On June 9 the company launched Claude Fable 5 alongside Claude Mythos 5, calling Fable its most powerful generally available model ever. Fable 5 is live today through the Claude API as claude-fable-5 and on Pro, Max, Team, and Enterprise plans — and it is already generally available in GitHub Copilot.

The headline is performance. On the agentic-coding benchmark SWE-Bench Pro, Fable 5 scores 80.3 — a wide margin over Claude Opus 4.8 at 69.2 and GPT-5.5 at 58.6 — and it became the first model to clear 90% on Hex's analytical benchmark. Anthropic also calls it a state-of-the-art vision model and points to long-horizon autonomy gains; one external tester reported reproducing a frontier physics result in 36 hours using about a third of the reasoning tokens GPT-5.5 needed over four days.

So what exactly is Fable? It is the public, guardrailed edition of Mythos — the model Anthropic withheld from the public for months because it was too good at finding and writing software exploits. The full Claude Mythos 5 stays restricted: its cyber safeguards are lifted only for vetted Project Glasswing partners, and its biology and chemistry safeguards only for select biomedical researchers. Fable is what is left once the most dangerous capabilities are fenced off.

That fencing is concrete. Fable 5 runs three classifier-based safeguards: for offensive cyber, for biology and chemistry, and for model distillation. When a request trips one, Fable refuses and quietly hands the query to Claude Opus 4.8 instead — Anthropic says more than 95% of Fable sessions never hit a guardrail at all. An external bug-bounty effort logged over 1,000 hours and turned up no universal jailbreaks, and the company is imposing a mandatory 30-day data-retention window on all Mythos-class traffic — even for enterprises with zero-retention contracts — to catch jailbreaks and false positives.

None of this is cheap. Fable 5 (and Mythos 5) list at $10 per million input tokens and $50 per million output tokens — less than half what Mythos Preview cost, but roughly double Claude Opus 4.8 and the priciest of the major frontier models. Anthropic is softening the blow at launch: from June 9 through June 22, Fable 5 is included at no extra cost on Pro, Max, Team, and seat-based Enterprise plans, after which continued use draws on usage credits.

The timing is striking. Fable arrives days after Anthropic publicly urged the industry to agree on a "coordinated brake pedal" for frontier AI, warning that systems may soon approach recursive self-improvement — and as the company races toward a potential record IPO alongside OpenAI. We flagged the Fable 5 rumors days ago; today they are confirmed, guardrails and all.