Anthropic's Fable 5.1 Release: AI Safety and Cost Advantages

Anthropic has unveiled the new release of its AI model Fable, version 5.1, featuring lower token costs and reduced safeguard restrictions. This version is paired with Mythos 5.1, a restricted model for cybersecurity and life sciences partners. Key highlights include Anthropic's zero data retention policy, enabling enterprise customers to run models on their own infrastructure without data outflow. Enterprise Frontier Safeguards, a high-privacy service, will be rolled out in the fall, allowing clients to control monitoring while still detecting misuse. The company assures customers that their data has never been inappropriately accessed. The new models set benchmarks in Terminal-Bench 4.0 and Humanity’s Last Exam, with Mythos 5.1 showing improved alignment compared to Opus 5. Fable 5.1 is now available unrestricted on cloud platforms and via the Anthropic API.

This marks a pivotal moment in AI safety, offering enterprise customers secure and cost-effective solutions while fostering innovation through open access. Anthropic's move positions it as a trusted partner for both businesses and researchers.