Anthropic Launches Claude Fable 5 and Restricted Mythos 5 for Cyber Defenders

Anthropic has released Claude Fable 5, its most capable general-purpose model to date, alongside Claude Mythos 5, a sibling model with safeguards lifted in select areas and restricted to a small group of cyberdefenders and infrastructure providers. Fable 5 leads on most capability benchmarks, with its biggest gains showing up on long-horizon agentic work in software engineering, knowledge work, vision, and scientific research. To manage risk, queries on sensitive topics like offensive cybersecurity get rerouted to the less capable Opus 4.8; the filter is tuned conservatively and fires on under 5% of sessions.

Mythos 5 is being deployed first through Project Glasswing, a collaboration with the US government, and Anthropic claims it has the strongest cybersecurity capabilities of any model in the world. Early results cited include Stripe compressing a multi-month Ruby migration into a day, protein design experts running roughly 10x faster on drug discovery workflows, and a Mythos-trained genomics model outperforming a recently published Science paper at one-hundredth the size. In blinded comparisons, scientists preferred Mythos’s molecular biology hypotheses over Opus-class outputs roughly 80% of the time.

Both models are priced at $10 per million input tokens and $50 per million output tokens, less than half the cost of Mythos Preview. Anthropic reports alignment metrics comparable to Opus 4.8 and says it plans to broaden Mythos access through a trusted-partner program as safeguards mature.