Hacker News
Anthropic 'Claude Mythos' Model Leaked via Unsecured Cache: 'Step Change' in Capabilities, Outperforms Opus in Cybersecurity and Coding
Fortune obtained leaked Anthropic blog drafts and nearly 3,000 unpublished assets from a publicly-accessible data cache revealing 'Claude Mythos' (also codenamed 'Capybara') — a new tier above Opus. An Anthropic spokesperson confirmed the model represents 'a step change' in performance, with significantly higher scores than Claude Opus 4.6 in software programming, academic reasoning, and cybersecurity. The model is in early-access testing with select customers and has no public release date.
Source
↳ Follow the thread