Benchmark
CyberGym Benchmark
What is CyberGym?
Real-world cybersecurity tasks; tool permissions and time limits are part of the result, not incidental to it.
- Evaluates
- security
- Openness
- unknown
- Importer
- Model reports
- Reported scores
- 2
Source provenance
- Original evidence https://openreview.net/forum?id=2YvbLQEdYt
CyberGym links
CyberGym results and reported scores
Model reports
| Model | Organization | Reported value | Reported | Evidence |
|---|---|---|---|---|
| DeepSeek-V4.1-Flash | DeepSeek | 88.1 | 2026-09-10 | evidence |
| Hy4 preview | Tencent | 78.4 | 2026-08-28 | evidence |
Scores are partitioned by the source that reported them and are never merged into a single cross-source ranking, because the sources measure different things and say so.
Which sources cite CyberGym?
- DeepSeek-V4.1-Flash DeepSeek, 2026-09-10
- Hy4 preview Tencent, 2026-08-28