Skills
Anthropic Claude Mythos Preview: 93.9% SWE-bench Verified, Autonomously Discovers 17-Year FreeBSD RCE (CVE-2026-4747)
Anthropic's unreleased Mythos Preview model scored 93.9% on SWE-bench Verified (vs 80.8% for Opus 4.6) and autonomously found and exploited a 17-year-old remote code execution vulnerability in FreeBSD NFS that grants root from unauthenticated internet access. Access is restricted to ~40 organizations via Project Glasswing, with Anthropic committing $100M in usage credits and $4M to open-source security organizations. This is the first frontier model to demonstrate near-complete autonomous engineering capability on well-specified tasks — and the first to independently discover a critical zero-day in production infrastructure.
↳ Follow the thread