Tool
CyberGym
- First Reported
- Mar 11, 2026
- Latest Reported
- Oct 21, 2026
Reported Context (10)
- For example, the Mythos model scored 83.1% on CyberGym's vulnerability reproduction benchmark as compared to 66.6% for Opus 4.6, a considerable leap in performance from Anthropic's prior leading model. Beazley Security Q1 2026 Report Highlights Supply Chain Attacks and Credential-Driven Ransomware
- The frontier question is whether a system can find a bug that demands deep reasoning, as we explored in our early work on the CyberGym benchmark. Microsoft FORGE Shares Lessons from AI-Assisted Vulnerability Research at Scale
- Atlas, a research initiative we shared earlier this year. At the time, Atlas held the number one spot on CyberGym and uncovered more than 200 vulnerabilities in widely used open source software. Wiz Previews AI SAST for Code Vulnerability Detection
- Yet it more than doubles GLM-5.2 on exploitation benchmarks and reaches state-of-the-art performance on CyberGym according to Z.ai’s evaluations. (Hugging Face) AI Cybersecurity Models Are Becoming More Interchangeable, Shifting Value to Workflows and Data
- The same day, a CyberGym API key and full exploitation recipe for a Modal-hosted customer workload is shared on the Artifactory board. OpenAI Says Reward-Hacking Agents Exploited Zero-Days in Hugging Face Intrusion
CVE (7)
Malware (6)
People (7)
Threat Actors (11)
MITRE ATT&CK (17)
Vendors (18)
Products (58)
Tools (12)
Industries (7)
Countries (2)
Note: Related entities, including threat actors, malware, CVEs, MITRE ATT&CK techniques, vendors, products, tools, countries, and industries, are shown when they appear in the same reporting. Their presence does not necessarily mean they were targeted, compromised, vulnerable, responsible for the activity, or directly involved in the incident.
