Anthropic Releases New Fable and Mythos Models with Cost Reductions and Enhanced Safeguards
On Tuesday, Anthropic launched two new versions of its advanced AI model: Fable and Mythos 5.1. The updates include cost reductions for tokens and fewer false-positive restrictions from the models' safety mechanisms. While Mythos 5.1 is limited to registered partners in cybersecurity or life sciences research, Fable 5.1 is available via cloud platforms and Anthropic's API. A key feature of this release is Zero Data Retention, allowing clients to run AI on their own infrastructure without data outflows. Anthropic also announced the rollout of a high-privacy service called Enterprise Frontier Safeguards in June, which will give users more control over monitoring for misuse while still maintaining oversight by Anthropic. The company emphasized that it has never trained models using enterprise data without explicit permission and reaffirmed its commitment to not doing so. The new models set records on several benchmarks like Terminal-Bench 4.0 and Humanity’s Last Exam, showcasing enhanced reasoning capabilities. Additionally, the release included three scientific findings generated by Anthropic's AI before public launch.
Topics: AI model release benchmark performance data privacy enterprise AI