Claude Mythos, Fable 5 & Mythos 5 — FAQ

It's Anthropic's Mythos-class model tier — the level above Opus — covering the Mythos Preview and the June 2026 editions Mythos 5 and Fable 5. See what is Claude Mythos.

They're the same underlying model. Fable 5 adds safety classifiers that reroute flagged requests to Opus 4.8; Mythos 5 has those lifted and is Glasswing-only. Full breakdown: Fable 5 vs Mythos 5.

Access is currently suspended under a US export-control directive. The live verdict is on Is Fable 5 available?, with the canonical view at anthropicmythos.ai/status.

$10 per million input tokens and $50 per million output tokens after the introductory window. Details: Fable 5 pricing.

Anthropic's restricted-access program for Mythos-class models — about 50 vetted organisations, weighted to cyberdefense and critical infrastructure. See Project Glasswing explained.

"Capybara" was an internal-rumor mislabel. The official tier name is Mythos-class. See the Capybara rumor, corrected.

1M tokens of input and up to 128k tokens of output — the same as Fable 5. See Mythos 5 context window.

Two reasons: Mythos 5 lacks Fable 5's safety classifiers (so it's Glasswing-only), and a June 12, 2026 export-control directive suspended both. See why is Mythos restricted.

Same underlying model; different safeguards. Fable 5 adds classifiers that reroute flagged requests to Opus 4.8. Mythos 5 has those classifiers lifted and is Glasswing-only.

Anthropic reports under 5% of sessions trigger a reroute, and you are told when it happens. See classifier fallback.

$10 per million input tokens and $50 per million output tokens after the introductory window. See Fable 5 pricing.

Access is currently suspended under a US export-control directive. Check the live status.

Yes — they share the same underlying model. The difference is safeguards: Fable 5 ships safety classifiers that reroute a small share of flagged requests to Opus 4.8, while Mythos 5 runs with those classifiers lifted and is Glasswing-only.

Because of a US export-control directive issued June 12, 2026 that suspended access. See why is Mythos restricted for the policy background.

The live status strip above and the canonical verdict at anthropicmythos.ai/status, both derived from Anthropic's official status page.

After the introductory window, $10 per million input tokens and $50 per million output tokens.

Yes. It was free for Pro, Max, Team, and Enterprise users from June 9 to June 22, 2026. Usage-based billing began June 23, 2026.

Fable 5 is about double Opus 4.8, which is priced at $5 input / $25 output per million tokens.

The $10/$50 figure is the reference rate. Mythos 5 is Glasswing-only, and the earlier Glasswing Preview used different pricing closer to $25/$125.

The unrestricted Mythos editions — the April 2026 Preview and Mythos 5 — with the safety classifiers lifted, rather than the safeguarded public Fable 5.

Roughly 50 were selected for the Mythos Preview, concentrated in cyberdefense and critical infrastructure.

No. It is selection-based with no public application, and Anthropic chooses participants directly.

Partly. Mythos 5 is gated to Glasswing by safety design because it runs without classifiers. But the current total suspension of both models is due to a separate June 12, 2026 export-control directive.

Anthropic intends to restore Fable 5 to subscription plans when capacity and clearance allow, once the export-control suspension permits.

No. Mythos 5 is the classifier-lifted model and is intended to remain available only to vetted organizations through Project Glasswing.

It is rerouted to Claude Opus 4.8 and answered there instead of by the full Mythos-class model, and the user is notified of the reroute.

Fewer than 5% of sessions trigger a reroute. The classifiers are deliberately narrow so most usage stays on the full model.

Not on Fable 5. The classifier-lifted version is Mythos 5, which is available only via Project Glasswing to vetted users.