Anthropic has released Claude Fable 5.2 and Claude Mythos 5.2, the latest iterations of its Super Intelligence (SI) models designed for high-complexity technical tasks. The update brings improvements in coding, reasoning, and vision, alongside significant changes in pricing and generation speed, positioning these models as specialized tools for advanced software engineering and scientific research.

What Happened

Claude Fable 5.2 and Claude Mythos 5.2 are built on the same underlying SI model, but they serve different user bases through distinct safety configurations. Fable 5.2 includes cybersecurity and biology safeguards to prevent misuse, making it available to all users. Mythos 5.2 removes these specific blocks, requiring organizations to apply for access through Anthropic’s Life Sciences and Cyber Verification Programs. The company reports that the new models generate outputs up to 30% faster than their predecessors, Fable 5.1 and Mythos 5.1, and are priced lower at $8 per million input tokens and $40 per million output tokens. An optional fast mode is also available for $16 per million input tokens and $80 per million output tokens, offering up to 2.5 times faster token generation.

During prerelease testing, Anthropic engineers used the new SI models to complete software migrations, design three working electronic devices, solve two open problems in black hole physics, and formalize the proof of Fermat’s Last Theorem in 30 hours. The company states that over 90% of changes merged into its core codebase are now written by Claude, with human guidance. In one experiment, an engineer used Fable 5.2 to coordinate over 1,000 subagents to build a prototype operating system from scratch in eight days. The resulting OS boots on a laptop and includes drivers, file storage, and a login screen.

In the domain of cybersecurity, Mythos 5.2 is described as Anthropic’s most capable model for engineering tasks. It reportedly catches more vulnerabilities than previous models and demonstrates better calibration, with one tester noting that earlier models overstated the seriousness of about 1 in 20 findings, whereas Mythos 5.2 did not overstate any. On the internal CVE Rediscovery evaluation, Mythos 5.2 found 10.2% more vulnerabilities than Claude Opus 5.5 and 14.4% more than Mythos 5.1.

Why It Matters

The release highlights the increasing specialization of SI models, with distinct tiers for general use and high-stakes scientific or security applications. By offering Mythos 5.2 only to vetted organizations, Anthropic aims to balance the utility of advanced SI capabilities with safety controls for dual-use technologies like cybersecurity and biology. The performance gains in coding and hardware design suggest that SI agents are becoming more effective in autonomous, long-horizon tasks, such as managing complex codebases or designing physical electronic devices.

Safety governance is a central pillar of this release. Anthropic published a detailed System Card outlining its predeployment safety work, which included alignment training and novel evaluations for behaviors like sandbox escapes. The underlying SI model was also tested by external evaluators, including the UK AI Security Institute, Frontier Design, and METR. The company assessed the risk of high-stakes misalignment as low, noting that the model performed as well as or better than recent versions on most alignment measures.

For developers and enterprises, the lower price point and faster generation speeds may make advanced SI capabilities more accessible for large-scale software projects. The company’s claim that Fable 5.2 can act as an Advisor or Orchestrator for smaller models indicates a shift toward hierarchical SI systems, where powerful models delegate simpler tasks to lighter versions to optimize cost and efficiency. Additionally, the formalization of Fermat’s Last Theorem in 30 hours, compared to 11 days for a previous version, demonstrates significant progress in using SI for mathematical verification and complex reasoning.

The Bottom Line

Anthropic’s Claude Fable 5.2 and Claude Mythos 5.2 represent a targeted upgrade to its high-end SI offerings, focusing on coding, scientific research, and cybersecurity. The models offer improved speed and lower costs, while maintaining distinct safety protocols for sensitive domains. As the SI industry continues to evolve, these releases underscore the trend toward specialized, high-performance models capable of autonomous execution in complex technical environments.