Anthropic PBC has confirmed the development of a new large language model (LLM) named Claude Mythos, which was unintentionally revealed due to a configuration error in its content management system. The company describes Claude Mythos as its most advanced model yet, currently undergoing pilot testing with select customers. This new model will feature a pricier fourth tier in its product lineup, surpassing the existing Opus variant, particularly in programming tasks and reasoning capabilities. Notably, the Capybara edition of Mythos excels in identifying cybersecurity vulnerabilities, prompting concerns among investors in firms like CrowdStrike and Palo Alto Networks. As cybersecurity threats evolve, Anthropic aims to equip organizations with tools to bolster code security proactively. Meanwhile, OpenAI Group has completed pretraining its own LLM, tentatively named Spud, set for release soon. As these advancements unfold, the competition in AI-driven security solutions intensifies.
Source link
Share
Read more