Mythos Proves Potent in Vulnerability Discovery, Less Convincing Elsewhere

Mythos appears to be as powerful as claimed at detecting software vulnerabilities, but its capabilities in other areas are more nuanced, according to SecurityWeek’s review. Independent benchmarking by XBOW found Mythos is extremely strong for source code audits and native-code vulnerability discovery, while its exploit validation and judgement can be uneven.

XBOW tests also showed Mythos excels when tested on “live + source” scenarios but is less convincing against source code alone, and that its judgment can be too literal yet sometimes misses true positives. The report notes that Mythos Preview is expensive, with Anthropic stating it will be 5x as costly as an Opus model, and XBOW concludes that while it isn’t best-in-class on all benchmarks, it remains powerful for finding candidate vulnerabilities from source code. Written by Kevin Townsend, 14 May 2026.