CWE-bench v2: More coverage, quality, and transparency

Collinear AI released CWE-bench v2, a public benchmark with 100 tasks covering 91 types of software security vulnerabilities. It evaluates AI agents on finding and fixing security flaws in real open-source code with programmatic, reproducible scoring. The benchmark aims for comprehensive coverage and fairness, with open-source tasks and leaderboards tracking model performance including open-weight and API-only models.