Versions

1 version · Metric: Vulnerability reproduction success

Versions of CyberGym
VersionReleasedModelsResults
CyberGym OriginalLatestJune 3, 2025910

Recently reported results: CyberGym Original

Latest version: CyberGym Original

Recently reported results for CyberGym Original
ModelScoreSource
Grok 4.7 (high)80.3%media.x.ai for Grok 4.7 (high) on CyberGym Original (opens in a new tab)Vulnerability reproduction successReported Model cardSpaceXAISelf-reportedRecord for Grok 4.7 on CyberGym Original
Grok 4.6 (high)79.7%media.x.ai for Grok 4.6 (high) on CyberGym Original (opens in a new tab)Vulnerability reproduction successReported Model cardSpaceXAISelf-reportedRecord for Grok 4.6 on CyberGym Original
GLM-5.3 (max)84.5%z.ai for GLM-5.3 (max) on CyberGym Original (opens in a new tab)Vulnerability reproduction successReported Launch postZ.aiSelf-reportedRecord for GLM-5.3 on CyberGym Original
Muse Spark 1.1 (high)59.0%ai.meta.com for Muse Spark 1.1 (high) on CyberGym Original (opens in a new tab)Vulnerability reproduction successReported Evaluation reportMetaSelf-reportedRecord for Muse Spark 1.1 on CyberGym Original
Muse Spark (high)43.5%ai.meta.com for Muse Spark (high) on CyberGym Original (opens in a new tab)Vulnerability reproduction successReported Evaluation reportMetaSelf-reportedRecord for Muse Spark on CyberGym Original