CyberGym
A public benchmark for evaluating AI models' cybersecurity capabilities, with a focus on assessing automated performance across the various stages of vulnerability discovery and exploitation chains.
Timeline (last 90 days)
Meta reports that Muse Wonder's cyber capabilities are 88.4% on CyberGym, 58.9% on ExploitBench, and 92 captured flags on ExploitGym.
On the CyberGym benchmark, Mistral Large 4 ranked first among all open-source models with a score of 82, ahead of MIMO, Grok, GLM 5.3, Kimi K3, and others.
On the CyberGym benchmark, Mistral Large 4 scored 82, ranking first among all open-source models and surpassing MIMO, Grok, GLM 5.3, Kimi K3, and others.
DeepSWE, CyberGym, and Automation Bench correspond respectively to evaluations for software engineering, cybersecurity, and automated agent tasks.
The release party claimed that DeepSeek-V4.1-Flash outperforms Opus 5 and GPT-5.6 Sol on three benchmarks: DeepSWE, CyberGym, and Automation Bench.
CyberGym is a public cybersecurity capability benchmark hosted on GitHub, used to evaluate AI performance on tasks such as vulnerability discovery and exploitation.
GLM-5.3 scored 84.5 on the CyberGym security capability benchmark.
CyberGym是专门评估AI模型在网络安全和代码安全审查领域能力的基准框架，涵盖漏洞检测、恶意代码识别、安全配置审计、依赖项风险评估等维度
All Facts (8)
GLM-5.3 scored 84.5 on the CyberGym security capability benchmark.
60%UnverifiedMeta reports that Muse Wonder's cyber capabilities are 88.4% on CyberGym, 58.9% on ExploitBench, and 92 captured flags on ExploitGym.
50%UnverifiedOn the CyberGym benchmark, Mistral Large 4 ranked first among all open-source models with a score of 82, ahead of MIMO, Grok, GLM 5.3, Kimi K3, and others.
50%UnverifiedOn the CyberGym benchmark, Mistral Large 4 scored 82, ranking first among all open-source models and surpassing MIMO, Grok, GLM 5.3, Kimi K3, and others.
50%UnverifiedThe release party claimed that DeepSeek-V4.1-Flash outperforms Opus 5 and GPT-5.6 Sol on three benchmarks: DeepSWE, CyberGym, and Automation Bench.
50%UnverifiedDeepSWE, CyberGym, and Automation Bench correspond respectively to evaluations for software engineering, cybersecurity, and automated agent tasks.
50%UnverifiedCyberGym is a public cybersecurity capability benchmark hosted on GitHub, used to evaluate AI performance on tasks such as vulnerability discovery and exploitation.
50%UnverifiedCyberGym是专门评估AI模型在网络安全和代码安全审查领域能力的基准框架，涵盖漏洞检测、恶意代码识别、安全配置审计、依赖项风险评估等维度
50%