Anthropic unveils its most powerful model, Claude Mythos, outperforming Opus 4.6.

Anthropic has released a preview version of its most powerful security vulnerability model, Claude Mythos, which outperforms Opus 4.6. The model can discover thousands of high-risk vulnerabilities and boasts attack capabilities surpassing most hackers. Due to its extremely high security risks, the model is not yet publicly available; it is only offered through "Project Glasswing" to giants like Amazon, Apple, and Google, as well as over 40 infrastructure companies, for defensive security hardening.