Mythos Proves Potent in Vulnerability Discovery, Less Convincing Elsewhere

"The model exhibits substantial strength in both native code vulnerability discovery and reverse engineering. In the reverse engineering tests, XBOW concluded Mythos is 'capable of triaging both its own results and competitor-model findings,' and t"

Mythos Preview shows stronger software vulnerability detection than existing models across providers. Testing indicates the model performs best when given both live system context and source code, while source-only probing yields weaker results. The findings align with the idea that design defects require higher-level understanding beyond code inspection. In judgment tasks, Mythos reduces false positives compared with predecessors but can miss true positives when evidence does not meet formal criteria, and it benefits from precise prompting. The model also demonstrates substantial capability in native code vulnerability discovery and reverse engineering, including triaging its own findings and competitor-model results. Additional evaluation covers assessment of native apps and visual acuity.

Read at SecurityWeek