Crypto
Home›Crypto›Regulation›Microsoft MDASH plus MAI-Cyber-1-Flash hits 95.95% on…
Microsoft MDASH plus MAI-Cyber-1-Flash hits 95.95% on CyberGym
Microsoft says MAI-Cyber-1-Flash handles up to 90% of MDASH workloads and cuts the combined setup cost by 50%.
Microsoft says its MDASH vulnerability hunting system, when paired with its new dedicated cybersecurity model MAI-Cyber-1-Flash, scored 95.95% on CyberGym, a benchmark that tests whether AI agents can reproduce known vulnerabilities across open-source projects.
According to Microsoft, the combined system outperformed several other models, including GPT-5.5 Cyber at 85.6%, Mythos 5 at 83.8%, GPT-5.6 Sol at 83.6%, and Gemini 3.5 Flash Cyber at 83.2%. Microsoft also said the scanner is in private preview through Microsoft Defender, where teams can review findings and generate proposed fixes.
Microsoft said MAI-Cyber-1-Flash is designed to process the bulk of tasks, handling up to 90% of the workload, while MDASH routes the hardest 10% of cases to GPT-5.4. The company added that using the combined setup is 50% cheaper than its current best MDASH configuration.
CyberGym measures performance by asking AI agents to reproduce 1,507 known vulnerabilities across 188 open-source projects and scoring by the percentage successfully reproduced in a controlled environment. Microsoft noted that its result was self-reported and was not on CyberGym's public leaderboard at publication time.