Anthropic is testing a new, highly capable AI model named Claude Mythos, described as a significant advancement over its previous Opus models. The existence of this model was revealed through a data leak, which exposed draft blog posts and internal documents detailing its capabilities in reasoning, coding, and cybersecurity. Anthropic confirmed the leak was due to a human error and is now carefully managing the release of Mythos through early access customers, acknowledging its potential cybersecurity risks. AI
Summary written by gemini-2.5-flash-lite from 1 source. How we write summaries →
IMPACT This new model, potentially named Capybara, sets a new benchmark for reasoning and coding, likely pushing competitors to accelerate their own frontier model development.
RANK_REASON Anthropic, a frontier AI lab, is testing a new, more capable model named Claude Mythos, as confirmed by the company following a data leak. [lever_c_demoted from frontier_release: ic=1 ai=1.0]