Claude Fable 5.2 and Mythos 5.2

Claude Fable 5.2 and Claude Mythos 5.2 are the latest upgrades to advanced AI models designed for tackling complex technical challenges. These models show significant improvements in coding, reasoning, and vision tasks. During pre-release testing, they have proven invaluable in software migrations, device design, physics problems, robotics evaluations, and even in formalizing the proof of Fermat's Last Theorem within 30 hours.

Claude Fable and Claude Mythos share the same underlying model. Fable comes with cybersecurity and biology safeguards to prevent misuse, making it accessible to all users. Organizations interested in beneficial applications in cybersecurity or biomedical research can apply for access to Mythos through Anthropic's Life Sciences and Cyber Verification Programs.

Users have noted that Fable 5.2 and Mythos 5.2 provide clearer communication than previous models, with outputs generated up to 30% faster. They also come at a lower price point, priced at $8 per million input tokens and $40 per million output tokens.

These models undergo rigorous safety testing to ensure alignment and safety as capabilities grow, including novel evaluations to identify potential issues like sandbox escapes. The team believes the risks of misalignment remain low, and the model's performance matches or surpasses previous models on most alignment measures. Fable 5.2's safeguards are as robust as those of Fable 5.1 but with fewer unnecessary blocks, and they've implemented the strongest protections against distillation attacks.

While Claude Opus 5.5 is recommended for everyday coding and knowledge tasks, Fable 5.2 and Mythos 5.2 excel in more ambitious, open-ended technical work. They can serve as Advisors or Orchestrators, delegating simpler tasks to smaller models. In terms of performance, Fable 5.2 and Mythos 5.2 outperform Opus 5.5 in most evaluated capabilities.

They also demonstrate superior engineering judgment and can produce high-quality results from ambiguous requests. One Anthropic engineer used Fable 5.2 to design a prototype operating system, coordinating over 1,000 subagents to create a fully functional system capable of booting on a laptop within eight days.

Claude Mythos 5.2, specifically, is the most capable model for cybersecurity engineering, catching more vulnerabilities than previous models and better reproducing and fixing them. In internal evaluations, it found 10.2% more vulnerabilities than Opus 5.5 and 14.4% more than Mythos 5.1. Mythos 5.2 also shows better judgment and calibration, catching fewer overstated findings compared to previous models.