Anthropic Launches Claude Fable – The First Public Release of Its Mythos Model

AI company Anthropic officially launched Claude Fable on Tuesday, marking the first public release of its most powerful model, Mythos—now equipped with strict safety guardrails. Users can access the model directly through Anthropic's Claude API and consumption-based enterprise plans. Previously limited to internal testing or select institutions, this public launch represents a significant step for Anthropic in opening up advanced AI capabilities.

At the same time, Anthropic also released Mythos 5—the latest version in the Mythos series—to authorized organizations. The company clearly stated that it will not use user data to train models. Instead, data is retained only for "defending against sophisticated attacks, including novel jailbreak techniques," and "identifying and reducing false positives." If widely adopted, this data retention policy could set a precedent in the industry, where access to more powerful models comes with mandatory data retention obligations in the name of safety.

In terms of performance, Anthropic revealed that scenarios requiring Fable to downgrade requests to the weaker Opus 4.8 model are extremely rare. Early data shows at least 95% of conversation sessions are handled entirely by Fable itself. This move aims to increase public accessibility while balancing model capability and safety risks. As AI models continue to grow more powerful, finding the optimal balance between openness and security will remain an ongoing challenge for the entire industry.