1. Mythos uses agent teams — multiple subagents coordinating directly with each other on segmented pieces of one task — which is a real architectural shift from single-agent Opus 4.6, not a marginal intelligence bump
2. It hit 83.1% on CyberGym vs Opus 4.6’s 66.6%, and converts 72.4% of identified vulnerabilities into working exploits autonomously — those are the only published numbers and they’re cybersecurity-specific, not general coding.
3. Anthropic says it will be deployed “when new safeguards are in place” — there is no announced public release date, no paid tier, and the Glasswing coalition is the only access path as of yesterday’s launch.