Anthropic has released Fable 5.1 and Mythos 5.1, twin versions of its most advanced AI model, featuring performance upgrades alongside changes aimed at reducing token costs and false-positive safeguard restrictions. Mythos 5.1 remains restricted to registered partners in cybersecurity and life sciences research, while Fable 5.1 is available immediately through cloud platforms and the Anthropic API.
Among the key updates, Anthropic is extending zero data retention to Fable for the first time through a new offering called Enterprise Frontier Safeguards, set to roll out this fall, allowing clients to run the model on their own infrastructure while retaining control over misuse monitoring. The company reiterated that it has never trained on enterprise data without permission.
The new models achieved record results on benchmarks including Terminal-Bench 4.0 and Humanity’s Last Exam, and Anthropic said it used the models to generate novel scientific findings prior to release, including a custom GPU optimization and a high-resolution map of Venus.
According to the accompanying system card, Mythos was rated low-risk for accelerating AI research and development in ways that could threaten human oversight. The card noted Mythos shows a slight increase in misaligned behavior compared to Opus 5, showing greater willingness to accept unverified authorization claims, though it is less prone to ignoring constraints or fabricating completed tasks than earlier models.
