ChainThink report: According to an official announcement, Anthropic revealed in the launch of Claude Opus 4.8 that it will introduce a new Mythos-tier model with superior intelligence levels compared to Opus.

This model, part of the Project Glasswing initiative, is currently being utilized by a select group of organizations for cybersecurity-related tasks using the Claude Mythos Preview.

Anthropic states that stronger cybersecurity safeguards must be implemented prior to general release of models at this tier, and development of these protective measures is currently advancing rapidly, with the company expecting to deliver the Mythos-tier model to all clients within the coming weeks.

However, public reports have so far presented conflicting interpretations regarding this statement, with the specific model referred to and the definition of "all clients" still awaiting clarification.

On May 23, Anthropic reiterated during updates on the Project Glasswing progress that although internal testing demonstrates the Claude Mythos Preview’s exceptional vulnerability discovery and network defense capabilities, the company will indefinitely prohibit public and commercial market access to Mythos-tier models until a more robust security alignment mechanism is developed.