safety benchmark
CyBench
CyBench is a suite of Capture-the-Flag (CTF) challenges measuring agentic cyber attack capabilities. It evaluates dual-use cybersecurity knowledge and measures the 'unguided success rate', where agents complete tasks end-to-end without guidance on appropriate subtasks.
Updated Aug 11, 2026
Models2
Model coverage2
MetricScore
EvidenceB
CyBench Ranking
Higher score ranks better on this benchmark.
| 01 | AN | 100.0% | 100.0% | 2 | C | |
| 02 | XA | 39.0% | 0.0% | 2 | C |
CyBench Score Distribution
A closer view of the leading scores on this benchmark.
CyBench Highlights
The leading models and scores on this benchmark.
What is CyBench?
What CyBench measures and how its scores work.
CyBench is a suite of Capture-the-Flag (CTF) challenges measuring agentic cyber attack capabilities. It evaluates dual-use cybersecurity knowledge and measures the 'unguided success rate', where agents complete tasks end-to-end without guidance on appropriate subtasks.
Scores are shown in ratio. This benchmark is not independently verified and has an evidence level of B.
- Family
- CyBench
- Modality
- text
- Primary category
- safety
- Score direction
- higher
- LLMBoard eligible
- No
- Evaluation key
- cybench|llm-stats-current
Benchmark scores retain their original unit. Overall score eligibility is shown separately.
FAQ
Common questions about CyBench.
Which model scores highest on CyBench?
Claude Mythos Preview is currently ranked first with 100.0%.
What does CyBench measure?
CyBench is a suite of Capture-the-Flag (CTF) challenges measuring agentic cyber attack capabilities. It evaluates dual-use cybersecurity knowledge and measures the 'unguided success rate', where agents complete tasks end-to-end without guidance on appropriate subtasks.
Is a higher score better?
Yes. Higher values rank better for this benchmark.
How many models are compared?
2 model results are currently shown.
Does this benchmark affect the overall score?
No. This benchmark is shown for reference but does not contribute to the overall score.