RELEASED · JUN 09 · 2026

Meet Claude Fable 5.
The Mythos-class model, made public.

Today we're releasing our most capable Claude ever — state-of-the-art on nearly every benchmark we run. Its sibling, Claude Mythos 5, runs the same brain with safeguards lifted in select areas, available to a small group of cyberdefenders and infrastructure providers through Project Glasswing.

What's actually new.

Six places Fable 5 jumps past Opus 4.8 — and where Mythos 5 goes further still.

Where Fable 5 sits.

The Claude family, ordered by raw capability. Mythos 5 leads; Fable 5 is the public face of the same model.

How Fable differs from Mythos.

Fable runs the Mythos brain behind a layer of classifiers tuned for public deployment.

01

Sensitive-topic fallback

Live classifiers route conversations involving cybersecurity, biology/chemistry, or model-distillation attempts down to Claude Opus 4.8.

02

Rare by design

Fallback triggers in fewer than 5% of sessions. The other 95%+ get the full Mythos-class response.

03

30-day retention

A new 30-day data retention window lets the safety team monitor edge cases and tune classifiers without becoming a long-term archive.

PUBLIC

Claude Fable 5

  • Available on the Claude API today
  • Pro / Max / Team / Enterprise plans
  • Safety classifiers active
  • 30-day retention for safety review
RESTRICTED

Claude Mythos 5

  • Project Glasswing partners only
  • Cyberdefenders & infrastructure providers
  • Safeguards lifted in select areas
  • Powers the scientific-research tier

Less than half the cost of Mythos Preview.

Frontier capability at a price that scales.

INPUT
$10 / 1M tokens
OUTPUT
$50 / 1M tokens
model_id claude-fable-5

Free on Pro, Max, Team and Enterprise plans through June 22, 2026. After that, requires usage credits.

Available now on the Claude API. Drop-in compatible with existing Opus integrations.

Less than half the price of Claude Mythos Preview.