Anthropic Unveils Claude AI Models for Enhanced Research

Anthropic has introduced two innovative AI models, Claude Fable 5.1 and Claude Mythos 5.1, aimed at transforming software engineering, corporate knowledge work, and scientific research. These models are touted to be the most proficient systems currently available in these fields.

Enhanced Features and Availability

While both models share the same architectural foundation, they differ significantly in terms of security measures. Claude Fable 5.1 is accessible to the general public, whereas Claude Mythos 5.1 is reserved for select professionals through Anthropic’s trusted access initiatives. These initiatives cater specifically to experts in cybersecurity and life sciences research.

This launch addresses three major issues raised by enterprise clients regarding previous Claude models: cost, data privacy, and restrictive content safeguards. In terms of pricing, Anthropic reveals that Fable 5.1 will be approximately 25% less expensive than its predecessor, and up to 45% cheaper for intensive tasks, thanks to a 75% reduction in cached input read costs.

Data Management and Privacy Enhancements

An important upgrade is the introduction of the Enterprise Frontier Safeguards (EFS) framework, which allows organizations to manage their data within their own cloud systems, ensuring zero retention on Anthropic’s servers while still maintaining security against misuse. This framework will be rolled out gradually across various platforms, including major cloud providers like AWS, Google Cloud, and Microsoft Azure.

Until the full deployment of EFS, eligible enterprises can adopt Fable 5.1 with a zero-retention policy. The model also reduces false-positive interventions in cybersecurity queries by 60%, enhancing its ability to identify software vulnerabilities for defensive purposes.

Performance and Compliance

Anthropic’s internal assessments show that Fable 5.1 surpasses its predecessor in coding, reasoning, and knowledge tasks, excelling in benchmarks such as Terminal-Bench 4.0 and GDPval-AA v2. Companies like Jane Street and MongoDB have already benefited from the model, solving long-standing engineering challenges.

Claude Mythos 5.1 expands into scientific fields, excelling in molecular biology by designing superior protein binders. It also created a new elevation map of Venus from old NASA data, ahead of future missions. These advanced capabilities are accessible through Anthropic’s verification programs, with plans to extend eligibility in the future.

Furthermore, Anthropic ensures compliance with the EU AI Act by embedding watermarking in outputs from models released after August 2026. Fable 5.1 is available immediately on Claude’s platforms and via API, continuing to innovate the field of AI research.