Anthropic just opened its top-tier Mythos class to the public with Claude Fable 5 — state-of-the-art capability wrapped in safeguards that fall back to Opus 4.8 on the riskiest queries. Here's what it does and how to get it.
Anthropic just opened the Mythos tier to the public
On June 9, 2026, Anthropic released Claude Fable 5 — the first Mythos-class model made safe for general use. Mythos sits a full tier above the Opus class, and until now it had only existed behind closed doors: Claude Mythos Preview shipped in April to a small circle of cyber defenders and critical-infrastructure providers under Project Glasswing. Fable 5 is the moment those capabilities reach the rest of us.
Alongside it, Anthropic shipped Claude Mythos 5 — the same underlying model, but with safeguards lifted for authorized partners in cybersecurity and biomedical research.
What "Mythos-class" actually buys you
Fable 5 exceeds every model Anthropic has previously released to the public, and it's state-of-the-art on nearly every benchmark tested. The pattern that matters: its advantage grows as tasks get longer and harder. A few highlights from launch partners:
- Software engineering. Stripe used Fable 5 to complete a 50-million-line Ruby migration in a single day — work the team estimated at two months. Cursor said it "opened up a class of long-horizon problems that were out of reach for earlier models," and it tops Cognition's FrontierCode evaluation.
- Knowledge work. Highest score on Hebbia's Finance Benchmark for senior-level reasoning, with big gains in document analysis and chart interpretation. One finance partner called it "the strongest finance-first model we've tested."
- Vision. Pulls precise numbers from scientific figures, rebuilds web apps from screenshots, and finished Pokémon FireRed using only raw game screenshots — no navigation aids.
- Memory & long context. Holds focus across millions of tokens. With persistent file-based memory, it tripled its own Slay the Spire performance versus Opus 4.8.
- Life sciences (Mythos 5). Roughly 10× faster protein-design workflows, novel molecular-biology hypotheses preferred by scientists 80% of the time over Opus-class models, and autonomous genomics work spanning 138 species.
Safety came first, deliberately
Fable 5 ships with three classifier-based safeguard systems covering cybersecurity, biology & chemistry, and distillation (blocking attempts to extract and replicate the model's capabilities in unauthorized jurisdictions). When a query crosses a high-risk line, Fable 5 quietly hands off to Claude Opus 4.8 instead of answering directly.
The tuning is conservative but light-touch: safeguards trigger in fewer than 5% of sessions on average. External red-teaming found no universal jailbreaks across 1,000+ hours of bug-bounty testing, and outside testers logged zero compliance across 30 jailbreak techniques on harmful cyberattack planning. All Mythos-class traffic carries a 30-day retention policy, is never used for training, and adds new access-logging protections.
How to get it
- API: Available now as
claude-fable-5— $10 / million input tokens, $50 / million output tokens — on the Claude API and consumption-based Enterprise plans. - Subscriptions: Included at no extra cost on Pro, Max, Team, and seat-based Enterprise from June 9–22. From June 23, it draws on usage credits, with full inclusion planned once capacity allows.
- Mythos 5: Restricted to Project Glasswing cybersecurity partners and selected biomedical researchers, with a broader trusted-access program coming soon.
The bigger picture
Anthropic spent months warning that frontier AI was getting too dangerous to release casually. Fable 5 is the counter-move: take the most capable thing they've built, wrap it in safeguards that fall back to a tamer model on the riskiest 5% of requests, and hand the other 95% to everyone. Whether that's the right balance is the debate of the year — but as of this week, Mythos-class capability is no longer hypothetical.
