Anthropic Launches Enhanced Mythos 5.2 and Fable 5.2 Models with Reduced Costs and Enterprise Privacy Features

Image: TechCrunch · Source

Anthropic has released updated versions of its AI models, Mythos 5.2 and Fable 5.2, boasting improvements in coding, reasoning, and vision, faster responses, lower costs, and new privacy safeguards tailored for enterprise clients.

On November 9, 2026, Anthropic unveiled the latest versions of its leading AI models, Mythos 5.2 and Fable 5.2, marking a notable advancement in their AI capabilities and enterprise readiness. According to TechCrunch, these models demonstrate enhanced performance in critical areas such as coding, reasoning, and vision, and offer clearer user communication.

One significant update is a reduction in Fable’s cost, now priced at $8 per million input tokens, making it more affordable than its chief competitor, GPT-6 Astra. Additionally, Fable 5.2 responds 30% faster across various effort levels, with an option for even quicker inference at doubled pricing.

Anthropic has addressed key enterprise concerns by introducing compatibility with zero data retention (ZDR) policies through its new Enterprise Frontier Safeguards (EFS) system. This innovation enables enterprises to meet strict privacy requirements by storing data within their own cloud infrastructure—supported on platforms including Amazon Web Services, Google Cloud, and Microsoft Azure—rather than Anthropic’s systems. The EFS solution allows customers to control data encryption keys and access policies, balancing privacy with effective automated monitoring for misuse detection.

The development of EFS involved extensive collaboration with over 100 enterprises across multiple regulated sectors such as finance, healthcare, and telecommunications. The system supports automated safety monitoring without requiring human review by Anthropic staff, addressing stringent regulatory and compliance demands.

The Fable 5.2 model has been quietly pre-released to select users and already contributed to significant scientific achievements, including solving two theoretical physics problems and formalizing Fermat’s Last Theorem. Although unconfirmed, a recent quantum computing research breakthrough utilized an undisclosed Anthropic model, possibly Fable 5.2.

This release follows an independent report by METR, which suggested Anthropic’s models might be approaching recursive self-improvement capabilities, potentially doubling AI progress speed. Anthropic emphasizes ongoing uncertainty around these findings and maintains that Mythos 5.2 currently lacks covert capacities harmful to users, despite improvements in reasoning obscurity.

Overall, Anthropic’s update reflects a step forward in AI model sophistication, operational efficiency, and enterprise-grade privacy controls, positioning these models as competitive and secure options for advanced AI applications in sensitive industries.