微软推出自研网络安全模型MAI-Cyber-1-Flash，但最棘手任务仍依赖OpenAI

Microsoft says it has closed the gap with frontier models in AI cybersecurity. The company is introducing MAI-Cyber-1-Flash , a compact security model built into its previously unveiled MDASH multi-agent system. The combination scores 96 percent on CyberGym, a benchmark measuring how well AI spots real security flaws in large codebases, putting it 12 points above Mythos and ahead of both Gemini and GPT. Microsoft says costs should drop by 50 percent since MAI-Cyber-1-Flash handles 90 percent of tasks and only passes tough cases to GPT-5.4. The model is based on the MAI-Thinking-1 line . Microsoft still relies on OpenAI for complex reasoning, but the setup fits its evolving role as an AI model orchestrator . That same shift has turned the Redmond company into an open-weights advocate after years of fueling Azure growth through exclusive OpenAI distribution. Microsoft is also launching Perception, an agent-based security system that monitors and mitigates threats in real time, citing its data advantage of over 100 trillion daily security signals and 1.6 million customers. Ad DEC_D_Incontent-1 Ad Subscribe to THE DECODER for ad-free reading, a weekly AI newsletter, our exclusive "AI Radar" frontier report six times a year, full archive access, and access to our comment section.