Anthropic has launched Fable and Mythos 5.1, paired versions of the company’s most advanced AI model, delivering performance gains along with a cheaper, less restrictive experience for developers. The new Fable release cuts token costs and reduces false-positive blocks triggered by the model’s built-in safeguards.
The two models serve different audiences. Mythos 5.1 remains limited to registered Anthropic partners working in cybersecurity or life sciences research. Fable 5.1, the unrestricted counterpart, is available starting today through cloud platforms and the Anthropic API.
Zero Data Retention and Enterprise Privacy
One of the most notable additions is Anthropic’s move toward zero data retention, which lets clients run the company’s models on their own infrastructure without data leaving their environment. This capability was previously unavailable for Fable because of security concerns.
A high-privacy tier called Enterprise Frontier Safeguards will begin rolling out to users in June. The system continues to monitor for misuse by both automated agents and human users, but customers will control how that monitoring is carried out. Anthropic also reassured customers that their data had not been improperly accessed, stating that it “has never trained on enterprise data without explicit permission, and never will.”
Benchmark Results and Scientific Findings
The new models set records across several benchmarks, including Terminal-Bench 4.0 for command-line coding tasks and Humanity’s Last Exam for general reasoning. Ahead of release, Anthropic used the models to produce three original scientific results, including a custom GPU optimization and a high-resolution map of Venus assembled from existing photography.
As with earlier launches, both models ship with a detailed system card that outlines their capabilities in plain terms. The document rates Mythos as “low-risk” for concerns tied to automated AI development, in which an AI system improves itself, a scenario some view as a trigger for loss of human control. According to the card, “its ability to accelerate internal AI R&D progress is in line with current trends.”
Alignment and Behavior Assessments
On general misbehavior, Mythos is slightly more prone to issues than Opus, possibly a byproduct of its expanded capabilities. The system card describes Mythos 5.1 as “a slight regression on overall misaligned behavior compared to Opus 5, and an improvement over Mythos 5 and Claude Sonnet 5.”
The card adds that the model “cooperates with human misuse and accepts unverifiable claims of authorization somewhat more readily than Opus 5, but it is less likely to ignore explicit constraints, hallucinate inputs, or falsely claim to have completed tasks than previous models.” Fable 5.1 is available on cloud platforms and through the Anthropic API as of today.
Source
Image: techcrunch.com
