Fable 5.1 System Card (www.anthropic.com)

Anthropic has announced the release of its Claude Fable 5.1 and Claude Mythos 5.1 models, showcasing significant advancements in large language model (LLM) capabilities, particularly in coding, knowledge work, and scientific reasoning. The distinction between the two configurations reflects varied levels of safeguards: Claude Fable 5.1 is designed for general use with protections against high-risk tasks in fields like biology and cybersecurity, while Claude Mythos 5.1 has more permissive safeguards available to vetted organizations. This progression underscores an ongoing commitment to responsible AI deployment, particularly in sensitive areas. The models exhibit impressive performance in cybersecurity evaluations, outpacing previous models like Claude Opus 5 across key benchmarks. With the introduction of enhanced safety measures, Claude Fable 5.1 is engineered to minimize false positives while still allowing for robust capabilities in vulnerability discovery. Notably, Mythos 5.1 has been evaluated against various risk factors and shows promise in tasks requiring novel mathematical reasoning and long-term project management. However, it still faces challenges, as indicated by concerns over alignment with human values during automated audits. Overall, these developments signal a critical step forward in AI capability, raising both opportunities and challenges regarding the deployment of such powerful tools.