Anthropic releases Claude Fable 5, its most powerful public model yet

On June 9, 2026, Anthropic released Claude Fable 5, the most capable model the company has ever made generally available. It is the first public model in a new "Mythos" class that, in Anthropic's words, "sits above our Opus class in capability."

A new tier above Opus

Until now, Opus sat at the top of Anthropic's lineup, with Sonnet and Haiku below it. Fable 5 adds a rung above all of them. It and a restricted sibling, Mythos 5, share the same underlying model — the difference is safety. Mythos 5 has safeguards lifted in some areas and is limited to a small group of cyberdefenders and infrastructure providers through Project Glasswing; Fable 5 is the version Anthropic has made safe for general use.

Anthropic says Fable 5 is "state-of-the-art on nearly all tested benchmarks" across software engineering, knowledge work, vision, and scientific research. One concrete data point: Stripe reported that Fable 5 "compressed months of engineering into days" on a migration of a 50-million-line Ruby codebase.

How the safety story works

To ship a frontier model to everyone, Anthropic added domain-specific guardrails. When a query touches cybersecurity, biology, chemistry, or model distillation, Fable 5 quietly defers to Claude Opus 4.8 instead. The company says these fallbacks trigger "in less than 5% of sessions on average," meaning at least 95% of Fable sessions run entirely on the new model. Anthropic also reports that an external bug bounty ran for "over 1,000 hours" without surfacing a universal jailbreak. A mandatory 30-day traffic retention requirement applies to all users.

The timing drew attention: as TechCrunch noted, the release came just days after Anthropic publicly warned that frontier AI is getting more dangerous.

Pricing and access

Fable 5 costs $10 per million input tokens and $50 per million output tokens — double the price of Opus 4.8 ($5 / $25), but less than half the price of the earlier Mythos Preview. Through June 22, it's included at no extra cost on Pro, Max, Team, and seat-based Enterprise plans; from June 23 it shifts to usage credits while Anthropic works to restore standard subscription access "as soon as possible." It's live now on the Claude API and reachable inside coding tools — /model fable in Claude Code and the model picker in Cursor.

Why it matters for people learning to build

For anyone building software with an AI assistant, the practical takeaway is simple: the ceiling on what a single agent run can attempt just went up again. Bigger refactors, longer multi-file changes, and gnarlier debugging sessions become more tractable. But the higher price is a real trade-off — most everyday coding still runs perfectly well on Sonnet or Opus, and reaching for the top tier on every task is an easy way to burn budget. The smart move is to keep Fable 5 in your back pocket for the genuinely hard problems, and let cheaper tiers handle the routine work.