[Anthropic Releases its Most Powerful Pre-IPO Models, Fable and Mythos 5.2: 1,000 AI Agents Collaborate to Build a System in 8 Days, Formalize Fermat's Last Theorem in 30 Hours] Beating AI News: Anthropic has released Claude Fable 5.2 and Mythos 5.2. These two models are now capable of handling engineering and research projects that previously required long-term, specialized teams, including developing operating systems from scratch, formalizing mathematical theorem proofs, and solving long-standing theoretical problems related to black holes. Anthropic enabled Fable 5.2 to develop an operating system starting from an empty folder. It mobilized over 1,000 sub-agents, working continuously for 8 days, autonomously dividing tasks, programming, testing, and reviewing code. The final system could boot on a laptop, possessing an independent kernel, disk and network drivers, a file system, a graphical window, and an installer. In another experiment, approximately 1000 Fable 5.2 agents completed the formalization of the existing proof of Fermat's Last Theorem in 30 hours, essentially converting the mathematical proof into code that a computer could verify step-by-step. Previously, Anthropic, using specialized tools and human researchers, took 11 days to complete a similar task. With the same configuration, Opus 5.5 completed only 26% of the task in about 30 hours. Fable 5.2 also helped solve two unsolved mysteries in black hole theory. One of them involved a complex calculation proposed in 2003 that had previously proven difficult to perform. Anthropic also had it independently design three electronic devices, including circuit boards, components, casings, and programs. Engineers at a partner company manufactured and assembled the devices according to the designs, and all three devices functioned correctly without requiring circuit modifications. In official evaluations, Fable 5.2 achieved a 49.1% success rate in the FrontierCode 2.0 programming test, surpassing Opus 5.5's 45.5%. Robot simulation testing improved from 27.6% in the previous generation to 64.9%. Mythos 5.2, geared towards cybersecurity and biological research, achieved 89.6% in known vulnerability rediscovery tests, compared to 78.3% in the previous generation. Fable and Mythos use the same set of model weights, with the main difference being security restrictions. Fable is open to all users, while Mythos has relaxed restrictions for cybersecurity and biological research, currently only available to approved institutions. Anthropic has also proactively limited the model's training progress on tasks such as AI chip core optimization and model architecture improvement, keeping these capabilities roughly at the Opus 5.5 level to reduce the risk of accelerating the development of stronger AI. API pricing is $8 per million input tokens and $40 per million output tokens, a 20% reduction from the previous generation, with a maximum 30% increase in generation speed. Anthropic still recommends prioritizing Opus 5.5 for general programming and everyday knowledge tasks, while Fable and Mythos are primarily geared towards complex, long-term technical projects. [Original Link]