CWE Bench
A benchmark for evaluating AI models' cybersecurity capabilities, assessing a model's ability to discover, validate, and remediate code vulnerabilities.
Core Facts
Timeline (last 90 days)
The CWE Bench score is 71%.
On the CWE Bench benchmark, Carbon achieved 71% and can turn verified vulnerabilities into runnable fix code.
Carbon scored 71% on CWE Bench, higher than Opus 5.5's 67%.
Gemini 4 Carbon achieved a score of 71% on the CW Bench cybersecurity test.
Argon is tied for first place on the CWE Bench leaderboard with two other models.
Argon is tied for first place on CWE Bench at 68% alongside GPT-6 Astra and Grok 4.7.
Argon tied for first place in the CWE Bench vulnerability remediation test with a score of 68%.
All Facts (6)
Gemini 4 Carbon achieved a score of 71% on the CW Bench cybersecurity test.
65%UnverifiedArgon tied for first place in the CWE Bench vulnerability remediation test with a score of 68%.
60%UnverifiedThe CWE Bench score is 71%.
50%UnverifiedOn the CWE Bench benchmark, Carbon achieved 71% and can turn verified vulnerabilities into runnable fix code.
50%UnverifiedCarbon scored 71% on CWE Bench, higher than Opus 5.5's 67%.
50%UnverifiedArgon is tied for first place on CWE Bench at 68% alongside GPT-6 Astra and Grok 4.7.
50%