Anthropic Launches Claude Fable 5.2 and Mythos 5.2

By Suad Seferi · 10 November 2026

Anthropic has introduced Claude Fable 5.2 and Claude Mythos 5.2, two advanced AI models designed to handle complex technical work. The announcement includes a demonstration in which more than 1,000 AI agents collaborated to build a functioning prototype operating system in just eight days. The experiment offers a glimpse of how artificial intelligence is moving beyond answering questions and generating code toward coordinating entire projects with limited human intervention. 1,000 AI agents working on one project According to Anthropic, an engineer instructed Fable 5.2 to develop an operating system starting from an empty folder. The model organised more than 1,000 specialised AI agents that worked continuously for eight days, writing software, testing components and reviewing one another's work. The result was a prototype operating system capable of running on a laptop, complete with file storage, network drivers, windows, a login screen and an installer. Unlike conventional AI coding assistants, which typically help developers complete individual tasks, this experiment demonstrated how multiple AI agents can coordinate work across a much larger software project. However, the demonstration took place under experimental conditions. Anthropic has not established that the system is suitable for everyday use or meets the reliability and security requirements of commercial operating systems. Beyond software development Anthropic also reported advances in scientific research and engineering. In one experiment, approximately 1,000 Fable 5.2 agents translated the mathematical proof of Fermat's Last Theorem into a computer-verifiable format in 30 hours, with minimal human intervention. The company also demonstrated AI-assisted hardware development, including three functioning electronic devices designed with limited human input. On the FrontierCode 2.0 software engineering benchmark, Fable 5.2 achieved 49.1%, compared with 45.5% for Claude Opus 5.5 and 42.8% for GPT-6 Astra, according to Anthropic's evaluations. These results suggest improvements in complex technical reasoning, although independent testing will be necessary to establish how consistently the model performs in real-world environments. Mythos 5.2 brings stronger cybersecurity capabilities Alongside Fable, Anthropic introduced Claude Mythos 5.2, which provides advanced capabilities for cybersecurity and biomedical research. The company says Mythos can identify and help repair software vulnerabilities more effectively than its predecessors. Unlike Fable, which is available to general users, Mythos requires organisational verification because some of its capabilities could also be misused. Fable 5.2 is available through Claude and supported cloud platforms. API pricing starts at $8 per million input tokens and $40 per million output tokens. Anthropic continues to recommend Claude Opus 5.5 for most everyday coding and knowledge tasks, positioning Fable as an option for more demanding technical projects. What this means for the future of AI The significance of Anthropic's announcement extends beyond the performance of another AI model. The operating system experiment demonstrates how AI systems can divide complicated work among specialised agents, coordinate their activities and produce a substantial technical result. For businesses and developers, this could eventually change how software is built, tested and maintained. Similar approaches may also accelerate scientific research and engineering. But greater autonomy introduces new challenges. When hundreds of AI agents contribute to the same project, identifying errors, verifying results and establishing accountability become increasingly important. Anthropic's demonstrations show what these systems can achieve in controlled experiments. The next challenge is proving that such capabilities can operate reliably, securely and economically in real organisations.