Microsoft Corp. announced a major expansion of its artificial intelligence models and agentic AI infrastructure at its Build developer conference in San Francisco. The centerpiece is Microsoft IQ, a unified intelligence layer now generally available, designed to make AI agents and Copilot context-aware and personalized for organizations.
Microsoft IQ includes Work IQ, which captures how users operate within Microsoft 365; Fabric IQ, which organizes structured business data on Microsoft Fabric; and Foundry IQ, which retrieves information from unstructured documents. The company also introduced Web IQ, a fast real-world grounding for agents using web search.
On the model front, the Microsoft Superintelligence Team released seven new in-house models. The flagship is MAI-Thinking-1, Microsoft's first reasoning model with 35 billion active parameters and a 128,000-token context window. The company claims independent raters preferred it over Anthropic's Sonnet 4.6.
Other new models include MAI-Image-2.5 for text-to-image and image-to-image generation, now live in PowerPoint; MAI-Transcribe-1.5 for high-accuracy transcription across 43 languages; MAI-Voice-2 with expanded language support; and MAI-Code-1, fine-tuned for GitHub and VS Code.
Microsoft also launched Codename MDASH, a multi-model agentic-security system that deploys over 100 agents to find exploitable bugs in code.
For business users, Microsoft announced Scout, a new personal agent built on OpenClaw and Work IQ. Scout remains attentive at all times, proactively handling meeting prep, scheduling conflicts, and routine tasks using local or cloud intelligence.