Anthropic's $100M security initiative reveals that expensive frontier models aren't necessary for finding critical software vulnerabilities—cheaper alternatives perform equally well.

Anthropic announced Mythos, a frontier model that discovered thousands of zero-day vulnerabilities across major operating systems and web browsers, including a 27-year-old bug in OpenBSD and a 16-year-old bug in FFmpeg.

Anthropic committed $100M in credits and $4M in donations to Project Glasswing, a consortium using Mythos to identify and patch critical vulnerabilities and construct autonomous exploits.

AISLE tested the same vulnerabilities against models costing 100x less than Mythos, with a 3.6 billion parameter model at $0.11 per million tokens successfully identifying the same critical 17-year-old FreeBSD remote code execution bug, suggesting expensive frontier models may be unnecessary for security research.