BREAKING
GPT-5.5-Cyber Tops Mythos 5
CyberGym pass@1 scores
GPT-5.5
81.8
Mythos 5
83.1
5.5-Cyber
85.6
0
vulnerabilities
0
projects
Defense-oriented workflow
1
Trace
↓
2
Validate
↓
3
Patch
↓
4
Evidence
Mixed reception
Praised
●
Fewer refusals on red teams
●
Automated patching tools
Concerns
●
Patches may add bugs
●
Gated access limits reach
Near-parity, more available
AI NEWS BLITZ
OpenAI says its new cyber model has overtaken Anthropic's Mythos 5 on CyberGym.