Microsoft Launches First Cybersecurity-Specific AI Model with 95.95% Benchmark Score

Microsoft has launched its first cybersecurity-specific AI model within MDASH, its multi-model vulnerability identification and remediation framework. The new configuration, combining MAI-Cyber-1-Flash and GPT-5.4, achieved a 95.95% score on the CyberGym benchmark. The company states this setup costs 50% less than its previous best-performing MDASH combination, which included GPT-5.4, GPT-5.4 mini, and GPT-5.3 Codex. Access to the new model is restricted to approved users. No specific dates, CVE IDs, or deployment locations were mentioned in the announcement.