Microsoft launches MAI-Cyber-1-Flash, a compact security model that cuts costs by 50% with MDASH multi-agent system

Microsoft has launched MAI-Cyber-1-Flash, a compact security model that handles 90 percent of vulnerability detection tasks on its own, reserving only the hardest cases for OpenAI's GPT-5.4. The model is embedded in Microsoft's previously announced MDASH multi-agent system, which coordinates multiple AI agents to find and fix security flaws. The company says the combination cuts costs by 50 percent compared to using frontier models for every task.

According to Microsoft, MAI-Cyber-1-Flash scores nearly 96 percent on the CyberGym benchmark, which tests how well AI identifies real security vulnerabilities in large codebases. That score is 12 percentage points above Mythos and ahead of both Gemini and GPT alone.

How MAI-Cyber-1-Flash works with OpenAI models

The model is based on Microsoft's MAI-Thinking-1 line and is designed for vulnerability identification, classification, and remediation guidance. Microsoft's MDASH system orchestrates multiple specialized agents: MAI-Cyber-1-Flash handles the bulk of tasks, while GPT-5.4 steps in only for complex reasoning that requires deeper analysis. Microsoft says this tiered approach is what enables the cost savings.

Key points about the system include:

• MAI-Cyber-1-Flash handles 90 percent of cybersecurity tasks autonomously.
• Only 10 percent of tasks are passed to OpenAI's GPT-5.4 for advanced reasoning.
• The combined MDASH system delivers world-class performance at 50 percent of the cost of leading models, per Microsoft.
• The model is part of Microsoft's broader push into AI-orchestrated security.

Implications for the security market and Microsoft's AI strategy

Microsoft also launched Perception, an agent-based security system that monitors and mitigates threats in real time. The company says it has a data advantage of over 100 trillion daily security signals and serves 1.6 million customers. As some industry executives express concern over the safety of new AI systems, protecting against those systems could become a major business opportunity for tech companies like Microsoft.

Microsoft's reliance on OpenAI for the hardest tasks highlights the company's evolving role as an AI model orchestrator rather than a pure model builder. That shift has turned Microsoft into an advocate for open-weight models after years of fueling Azure growth through exclusive OpenAI distribution. The launch of MAI-Cyber-1-Flash and Perception comes as Microsoft continues to invest heavily in AI for cybersecurity, a market that is rapidly expanding as threats become more sophisticated. The company plans to integrate these models into its existing security offerings for enterprise customers later this year.