Mythos Preview appears to be exceptionally effective at finding software vulnerabilities, with XBOW confirming Anthropic’s claim that it outperforms other models in source code audits. However, its performance is more mixed in judgment, exploit validation, and cost efficiency, where it can be overly literal and is not always best-in-class. #Anthropic #MythosPreview #XBOW #Opus4.6 #GPT5.5
Keypoints
- XBOW confirmed that Mythos Preview is a major step up for finding vulnerabilities in source code.
- The model performs better when testing both live systems and source code together.
- Mythos shows strong reverse engineering and native-code vulnerability discovery capabilities.
- Its judgment can be too literal, sometimes missing true positives or overstating relevance.
- Mythos is powerful, but its high cost limits its efficiency compared with cheaper alternatives.