Anthropic Announces Claude Fable 5.2 and Mythos 5.2
Anthropic has announced Claude Fable 5.2 and Claude Mythos 5.2, featuring massive agentic multi-agent scaling designed to tackle highly complex technical tasks. A notable achievement highlighted in the release is a team of 1,000 Fable 5.2 agents fully formalizing Fermat's Last Theorem in just 30 hours with minimal human intervention. This release represents a major leap forward in collaborative agentic AI, demonstrating that large groups of autonomous models can coordinate and solve grand scientific challenges previously thought impossible. It sets a new benchmark for AI performance in specialized domains like coding, advanced mathematics, and finance. Claude Fable 5.2 is available for general use with built-in safeguards against high-risk dual-use domains like biology and cybersecurity, while the Mythos 5.2 configuration targets specialized high-end challenges. Technical documentation notes specific migration requirements, such as ensuring each thinking block aligns with the preceding conversation history and system prompt.
## BACKGROUND
Formalizing a mathematical theorem involves translating standard mathematical proofs into a rigorously structured computer language that automated proof assistants can mathematically verify. Multi-agent systems use groups of autonomous AI agents working together, often dividing large projects into smaller sub-tasks, to accomplish complex goals that a single model cannot easily handle alone.