Anthropic Unveils Claude Fable 5.2 With Major Improvements in AI Reasoning
The upgraded AI model promises faster performance, stronger coding abilities and major advances in scientific research, while its more restricted Mythos version targets cybersecurity and biomedical applications.
Anthropic has introduced Claude Fable 5.2, an upgraded artificial intelligence model designed to handle complex technical challenges, advanced coding and demanding research tasks. Announced on November 9, 2026, the new model delivers improvements in reasoning, visual understanding and long-running AI operations. According to Anthropic, it generates responses up to 30 per cent faster than its predecessor, Claude Fable 5.1, while also offering lower pricing. The company says the upgrade is designed to help developers and businesses complete complicated projects with less human supervision.
One of the major improvements is its ability to coordinate multiple AI agents and work independently on large software projects. Anthropic revealed that an experimental version of Fable 5.2 coordinated more than 1,000 AI agents over eight days to build a working operating system from scratch. The system reportedly included hardware drivers, file storage, windows, a login screen and an installer. The company also said more than 90 per cent of changes merged into its core codebase are now written by Claude, with human guidance.
Beyond software development, Anthropic highlighted the model’s achievements in mathematics, physics, hardware engineering and robotics. In one experiment, approximately 1,000 Fable 5.2 agents completed a computer-verified formalisation of Fermat’s Last Theorem in 30 hours, with minimal human intervention. The company also reported that the model helped solve two previously open theoretical physics problems involving black holes and designed three working electronic devices. These results suggest that advanced AI systems could increasingly support specialised scientific and engineering work.
Alongside Fable 5.2, Anthropic also launched Claude Mythos 5.2, which shares the same underlying model but has fewer restrictions for approved cybersecurity and biomedical research organisations. Mythos access will remain limited to organisations accepted into Anthropic’s verification programmes, while Fable 5.2 is available to the general public through Claude and major cloud platforms. The standard model costs $8 per million input tokens and $40 per million output tokens, with an optional faster mode available at higher prices. The launch strengthens Anthropic’s position in the increasingly competitive market for advanced AI systems.
