Anthropic Announces New Fable and Mythos 5.1 AI Model Versions
Anthropic has launched the Fable and Mythos 5.1 versions of its advanced AI models, offering lower costs and flexible constraints.
Anthropic has released the new Fable and Mythos 5.1 versions to users, enhancing AI model performance, lowering token costs, and reducing false positive constraints.
New Models and Version Details
Anthropic has officially announced Fable and Mythos 5.1, which are twin versions of the company's most advanced AI model. Alongside performance upgrades, the new Fable release aims to lower token costs and reduce false positive restrictions resulting from the model's safety measures.
Access Conditions and Distribution Channels
Like the previous Mythos model, the Mythos 5.1 version will be available exclusively to registered Anthropic partners conducting cybersecurity or life sciences research. The unrestricted version, Fable 5.1, is accessible via cloud platforms or through the Anthropic API.
Privacy Policies and Data Retention
One of the most significant changes is the adoption of a Zero Data Retention policy by Anthropic. This enables customers to run Anthropic models on their own infrastructure without experiencing data egress.
New Security Measures and Monitoring
Enterprise Frontier Safeguards, a high-privacy service previously unavailable for Fable due to security concerns, will be rolled out to users in June. The system will still monitor for misuse by agents or human users, but customers will control how that monitoring is performed.
Performance Tests and Scientific Findings
The new models broke records on various benchmark tests, including Terminal-Bench 4.0 and Humanity's Last Exam. Anthropic also shared three new scientific findings prior to the models' release, including a dedicated GPU optimization and a high-resolution map of Venus.