Opus 4.6 Model Significantly Improves Detection of High-Severity Vulnerabilities

The Opus 4.6 model is significantly better at detecting high-severity vulnerabilities compared to previous models. Unlike traditional fuzzers, Opus 4.6 analyzes and reasons about code in a manner similar to a human researcher, identifying problematic patterns and understanding the logic of the code to find specific inputs that break it. During tests on well-tested codebases, Opus 4.6 discovered high-severity vulnerabilities, some of which had remained undetected for decades.