Anthropic launches Claude Fable 5 and Mythos 5 with tiered safety protocols

Anthropic has officially launched two new Mythos-class artificial intelligence models: Claude Fable 5, designed for broad public accessibility, and Claude Mythos 5, which features lifted safety restrictions for specific institutional partners. The release marks a significant shift in the company’s approach to deploying high-capability models, balancing rapid market availability with stringent security protocols for sensitive applications.

Claude Fable 5 is now available to all users via the Claude API and Enterprise plans. While the model demonstrates state-of-the-art performance in software engineering, knowledge work, and scientific research, it operates with conservative safety filters. These safeguards automatically redirect queries related to cybersecurity, biology, chemistry, and model distillation to Claude Opus 4.8, a slightly less capable but safer alternative. Anthropic noted that these filters trigger in less than five per cent of sessions, prioritising risk mitigation over seamless user experience.

For a more unrestricted experience, Claude Mythos 5 is deployed through Project Glasswing, a private initiative in collaboration with the US government. This version is currently accessible to cyber defenders and critical infrastructure providers, offering the strongest cybersecurity capabilities in the market. Anthropic plans to expand this access to approximately 150 new organisations across more than 15 countries, alongside a future trusted-access program for biology researchers.

The pricing structure for both models is set at $10 per million input tokens and $50 per million output tokens. This rate is double that of Claude Opus 4.8 but represents less than half the cost of the previous Claude Mythos Preview. Anthropic stated that the joint launch aims to bring advanced AI capabilities to users quickly and safely, with Fable 5 expected to see high demand on consumption-based plans.

Early testing highlights substantial efficiency gains, with companies like Stripe reporting that Fable 5 compressed months of engineering work into days. In life sciences, internal testing using Mythos 5 accelerated drug design processes by ten times, with the model autonomously identifying strong candidates for protein targets. Anthropic intends to publish these genomic and molecular biology results in the coming months, underscoring the dual-use potential and risks inherent in such powerful systems.