Summary
Anthropic has officially launched Claude Fable 5, the first publicly available model from its new “Mythos” capability class, with general availability confirmed on June 11, 2026. The model delivers state-of-the-art performance across software engineering, knowledge work, vision, and scientific research benchmarks, representing a significant leap over the previous Opus class.
Alongside Fable 5, Anthropic introduced Claude Mythos 5 — an unrestricted version available only through a limited trusted-access program called Project Glasswing, initially targeting cyberdefenders and infrastructure providers. The key difference: Fable 5 automatically routes certain high-risk queries (cybersecurity, biology, chemistry, distillation) to the less capable Claude Opus 4.8, while Mythos 5 has these safeguards removed.
Both models are priced at $10 per million input tokens and $50 per million output tokens — less than half the cost of the earlier Mythos Preview. Fable 5 is currently included in Anthropic subscription plans (Pro, Max, Team, Enterprise) at no extra cost until June 22, after which usage shifts to per-token API billing.
Source
Anthropic Documentation · The Guardian · Fast Company · Forbes
Commentary
The dual-model approach — public Fable 5 with guardrails, restricted Mythos 5 without — is Anthropic’s attempt to thread the needle between capability and safety. It’s a pragmatic solution, but the automatic rerouting of cybersecurity queries to a weaker model has already drawn criticism from security researchers who argue it hampers legitimate defensive work.
The timing is strategic: with an IPO filing already submitted and Dario Amodei publicly calling for FAA-style AI regulation, Anthropic is positioning itself as the “responsible AI” company that still ships frontier capabilities. Whether that narrative holds depends on how the Mythos 5 trusted-access program evolves — and whether competitors like OpenAI and Google adopt similar tiered access models or use the restrictions as a competitive wedge.