New: connect Claude & other AIs to GenAIList over MCP — research the catalog and contribute to the shared knowledge base. Learn how →
CWE-Bench-Java benchmark
AI model leaderboard for the CWE-Bench-Java benchmark. Compare how large language models score on CWE-Bench-Java, see the full ranking, and understand what this AI benchmark measures. Claude Mythos Preview currently leads with 71.5%. Real-world Java vulnerability benchmark used in Anthropic's Project Glasswing cybersecurity evaluation.
Leaderboard
| # | Model | Organization | Score | Variant | Source |
|---|---|---|---|---|---|
| #1 | Claude Mythos Preview | Anthropic | 71.5% | official ↗ |
Frequently asked questions about CWE-Bench-Java
What is the CWE-Bench-Java benchmark?
Real-world Java vulnerability benchmark used in Anthropic's Project Glasswing cybersecurity evaluation.
How is the CWE-Bench-Java benchmark scored?
CWE-Bench-Java is scored using the detection-rate (%) metric, where a higher score is better. The maximum achievable score is 100.000. GenAIList aggregates reported scores from model providers and papers into a single ranked leaderboard.
Which AI model scores highest on CWE-Bench-Java?
As of the latest reported scores on GenAIList, Claude Mythos Preview achieves the highest result on CWE-Bench-Java with a score of 71.5%.
Is a higher CWE-Bench-Java score better?
Yes. On CWE-Bench-Java a higher score indicates better performance, so models near the top of the leaderboard are the strongest.