Anthropic has officially launched Claude Fable 5, its most advanced artificial intelligence model yet made available for general use. This release is accompanied by Claude Mythos 5, an identical but less restricted version designed for specialized partners in high-stakes fields. The launch represents a significant step in deploying frontier AI capabilities while implementing robust safety protocols to mitigate potential misuse.
A New Benchmark in AI Capabilities
Claude Fable 5 demonstrates state-of-the-art performance across a wide range of demanding tasks, setting new industry benchmarks in multiple domains. Early testing partners, such as Stripe, reported compressing months of complex software engineering work into a matter of days. The model also excels in knowledge work, vision-based analysis, and scientific research, consistently outperforming previous models on complex, long-horizon problems.
Prioritizing Safety with a Guardrail System
To address the risks associated with such a powerful model, Anthropic has integrated a unique safeguard system into Fable 5. For queries related to sensitive topics like cybersecurity, advanced biology, or chemistry, the model automatically defers to Claude Opus 4.8. This fallback mechanism, which triggers in less than 5% of sessions, is designed to prevent misuse while maximizing public access to its core strengths.
The Unrestricted Mythos 5 and Trusted Access
In parallel, Claude Mythos 5 offers the same underlying technology as Fable 5 but without the same level of safeguards for specific domains. This version is initially being deployed through Project Glasswing to a select group of cyberdefenders and critical infrastructure providers. Anthropic plans to expand availability through a trusted access program for both cybersecurity and biomedical research organizations needing unrestricted capabilities.
Industry Reception and New Policies
Early feedback from industry partners has been overwhelmingly positive, with companies like GitHub and Cognition praising its advanced coding and reasoning abilities. Alongside the launch, Anthropic introduced a mandatory 30-day data retention policy for business customers to monitor for sophisticated threats. This safety-focused policy, which will not be used for model training, could establish a new industry standard for frontier models.
Pricing and Availability
Both Fable 5 and Mythos 5 are priced at $10 per million input tokens and $50 per million output tokens, a significant increase over previous models. Fable 5 is immediately available through the Claude API and consumption-based enterprise plans, with a phased rollout for subscription users. This premium pricing structure reflects its advanced capabilities and the high demand anticipated by the company for its most powerful offering.
Anthropic's dual release of Claude Fable 5 and Mythos 5 marks a pivotal moment in the responsible deployment of advanced AI. By offering a public version with built-in guardrails alongside a restricted version for vetted experts, the company is navigating the complex challenge of balancing innovation with safety. This strategic approach aims to democratize access to powerful tools while actively managing the inherent risks of frontier technology.