Anthropic Debuts Claude 3.5 Sonnet for Advanced AI Capabilities

Key Takeaways

• Anthropic has launched two advanced AI models, Claude Fable 5.1 and Mythos 5.1, designed for enterprise and scientific applications.
• The new models address key enterprise concerns regarding cost, data privacy, and overly restrictive content safeguards.
• Fable 5.1 is generally available, while Mythos 5.1 is restricted to vetted professionals for high-stakes research and cybersecurity tasks.
• Significant improvements include reduced false positives for cybersecurity queries, enhanced data privacy through client-side storage, and lower operational costs.

Anthropic Unveils Next-Generation Claude Fable 5.1 and Mythos 5.1 AI Models

Anthropic has introduced its latest advancements in artificial intelligence with the release of Claude Fable 5.1 and Claude Mythos 5.1. These new frontier AI models are positioned as the most sophisticated systems developed by the company to date, tailored to enhance software development, enterprise knowledge management, and scientific research.

While sharing a common architectural foundation, Fable 5.1 and Mythos 5.1 diverge in their application and access. Fable 5.1 is broadly accessible to all users, whereas Mythos 5.1 is reserved for verified professionals participating in Anthropic’s trusted access programs. Access to Mythos 5.1 is specifically granted to cybersecurity defenders and life sciences researchers, with permissions carefully configured for their specialized needs.

This release directly responds to persistent feedback from enterprise clients, who frequently highlighted three critical areas for improvement in previous Claude iterations: operational costs, data privacy, and the stringency of content safeguards.

Cost Efficiency and Data Privacy Enhancements

Regarding pricing, Anthropic says Fable 5.1 offers a notable reduction in costs. Typical workloads are expected to see a 25 percent price decrease compared to Fable 5, with heavily agentic tasks potentially realizing savings of up to 45 percent. This substantial reduction is primarily driven by a 75 percent cut in cached input read prices, now set at $0.25 per million tokens. Standard input and output token pricing remains constant at $10 and $50 per million tokens, respectively.

Data handling has undergone a significant transformation with Anthropic’s new Enterprise Frontier Safeguards (EFS) framework. This innovation enables organizations to store their data exclusively within their own cloud infrastructure, bypassing Anthropic’s servers entirely. This architecture ensures zero-data-retention-level privacy for customers while maintaining robust defenses against adversarial misuse.

The EFS framework will be progressively rolled out beginning this fall across various Anthropic offerings, including Claude Code, Claude Enterprise, the Claude Platform, and key cloud partners such as AWS, Google Cloud, and Microsoft Azure. In the interim, eligible enterprise clients can immediately leverage Fable 5.1 under a zero-retention data policy.

Refined Safeguards and Performance Benchmarks

Anthropic reports a 60 percent reduction in false-positive interventions for cybersecurity-related queries with Fable 5.1. This improvement is partly due to the model’s new capability to identify software vulnerabilities for defensive purposes, though it remains restricted from generating exploits. Higher-risk dual-use applications, such as penetration testing and binary vulnerability scanning, are automatically rerouted to Anthropic’s more advanced Opus-class models.

Internal evaluations by Anthropic demonstrate Fable 5.1’s superior performance across critical benchmarks for coding, agentic reasoning, and knowledge work. These include Terminal-Bench 4.0, GDPval-AA v2, and Humanity’s Last Exam, where the new model consistently outpaced its predecessor.

Early enterprise partners, including Jane Street, Cognition, MongoDB, Datadog, and Millennium, have attested to the model’s capacity to resolve complex problems that had long challenged human engineering teams. One notable success involved Fable 5.1 debugging a rare production crash that had eluded human experts for nearly five years.

Mythos 5.1 extends its research prowess into scientific domains. In molecular biology experiments, the model successfully designed high-affinity protein binders, outperforming leading entries in Adaptyv Bio’s design competitions. It achieved nearly a 50 percent hit rate, significantly higher than the typical industry benchmark of 10 to 15 percent.

Additionally, Fable 5.1 independently generated a new high-resolution elevation map of Venus, utilizing decades-old NASA Magellan radar data. This achievement provides valuable insights ahead of upcoming ESA and NASA missions to the planet.

Mythos 5.1’s advanced capabilities in biology and cybersecurity remain under restricted access, governed by Anthropic’s Cyber Verification Program and Life Sciences Verification Program. These programs were developed in collaboration with the U.S. government, with plans to expand eligibility over time.

Anthropic has also confirmed its adherence to the EU AI Act’s Code of Practice. All model outputs released after August 2, 2026, will feature embedded invisible watermarking. A corresponding detection API is currently in private preview for regulators and researchers. Fable 5.1 is immediately accessible on Claude’s platforms and via its API, identified as claude-fable-5-1.

Disclaimer: HackersRadar reports on cybersecurity threats and incidents for informational and awareness purposes only. We do not engage in hacking activities, data exfiltration, or the hosting or distribution of stolen or leaked information. All content is based on publicly available sources.