SEC-bench Pro
An evaluation suite oriented toward cybersecurity capability, focused on discovering vulnerabilities in real codebases and examining models' ability to recognize defect patterns.
Core Facts
Timeline (last 90 days)
Gemini 4 Carbon achieved 88.2% on the SEC-bench Pro benchmark.
SEC-bench Pro is a specialized benchmark in the security domain, with tasks designed around real defense scenarios such as vulnerability identification, malicious code analysis, and patch generation.
Carbon achieved state-of-the-art results on all three cybersecurity sub-benchmarks—SEC-bench Pro vulnerability discovery, ExploitGym vulnerability exploitation, and CWE-bench patch repair—at 88.2%, 54.8%, and 71.0%, respectively.
All Facts (3)
Gemini 4 Carbon achieved 88.2% on the SEC-bench Pro benchmark.
65%UnverifiedSEC-bench Pro is a specialized benchmark in the security domain, with tasks designed around real defense scenarios such as vulnerability identification, malicious code analysis, and patch generation.
50%UnverifiedCarbon achieved state-of-the-art results on all three cybersecurity sub-benchmarks—SEC-bench Pro vulnerability discovery, ExploitGym vulnerability exploitation, and CWE-bench patch repair—at 88.2%, 54.8%, and 71.0%, respectively.
50%