Anthropic Claude Fable 5 Launch

Anthropic released Claude Fable 5 on June 9, 2026, a public Mythos-class AI model for enterprise customers and paid subscribers. The model removes Mythos’ offensive cybersecurity capabilities, routes dangerous queries to an older Opus model, and is reported to cost roughly twice as much as the prior flagship.

Enterprise Launch and Pricing

Claude Fable 5 is positioned as Anthropic’s new flagship large model, optimized for complex, long-context, multi-turn enterprise workloads. The company targets knowledge-intensive sectors such as finance and technology, where extended reasoning and multi-step analysis are essential.

The model became immediately available through Anthropic’s APIs and user interfaces after a period of restricted testing. Initial reports indicate a significant price premium for Fable 5, though specific per-token rates and contract terms remain undisclosed. This pricing will be a key factor in adoption and revenue, as enterprise customers weigh the higher capability against cost.

Safety Guardrails and Query Routing

Fable 5 shares its architecture with Claude Mythos 5 but includes additional safety measures to limit dual-use capabilities—functions that could be used for both beneficial and harmful purposes. Anthropic removed the Mythos preview’s offensive cybersecurity features and implemented guardrails barring use in risky domains such as cybersecurity exploitation and biological threats.

When safety filters detect high-risk prompts—examples include exploiting software vulnerabilities, writing malware, penetration testing, and designing biological weapons—requests are routed to an older Claude Opus model. This routing preserves access to a general-purpose assistant while restricting Mythos-class behavior in sensitive areas.

The public launch follows an April 2026 restricted Mythos preview under Project Glasswing, which demonstrated the model’s ability to identify critical software vulnerabilities. That preview raised substantial concern among cybersecurity experts and financial institutions, prompting Anthropic to withhold full public access until additional safeguards were in place.

Anthropic intends to commercialize Mythos-class capabilities under stricter safety regimes, using Fable 5 as a template for high-capability but guardrailed public models. The full Mythos preview remains limited to vetted institutional partners as the company evaluates safety outcomes. This approach aims to balance market demand for advanced AI with tighter operational controls and risk management.