Claude Fable 5: Anthropic ships its first Mythos-class model to everyone — with a catch on every risky query

Anthropic released Claude Fable 5 on June 9 — the first model from its Mythos-class tier (above Opus) cleared for general use. Anthropic calls it state-of-the-art on nearly every benchmark it tested, and the launch-day partner quotes back that up: Hex reported the first-ever 90% on its core analytics benchmark, and Rakuten credited Fable's self-validation for making "highly autonomous operations possible."

The mechanism worth understanding before you build on it: Fable 5 ships with classifiers watching for cybersecurity, biology/chemistry, and model-distillation requests. When one trips, your query is silently answered by Claude Opus 4.8 instead, with a disclosure. Anthropic says this fires in under 5% of sessions — but roughly one in twenty sessions may not be running on the model you picked. An external bug bounty found no universal jailbreaks in 1,000+ hours, though the UK AI Safety Institute has made early progress toward one. New for everyone: Mythos-class traffic now carries a mandatory 30-day data retention, even for enterprises with zero-retention agreements.

The practical details: API access via claude-fable-5 is live today at $10/$50 per million tokens — double Opus 4.8. Fable 5 is included free on Pro, Max, Team, and seat-based Enterprise plans only through June 22; on June 23 it moves to usage credits. The sibling Mythos 5 (cyber safeguards lifted) stays gated to government and critical-infrastructure partners via Project Glasswing.

Sources: Anthropic, TechCrunch, VentureBeat