Microsoft launches Deep Research public preview in Azure AI Foundry
Microsoft rolled out the public preview of Deep Research in Azure AI Foundry. This new API and SDK offering plugs OpenAI’s advanced agentic research into Azure’s enterprise platform.
Deep Research lets developers build agents that plan, analyze, and synthesize web info at scale. It automates complex research tasks with traceable, source-backed outputs. Agents can be composed into multi-step workflows alongside other Azure AI Foundry tools. 
The launch targets enterprises wanting more than simple chatbots. It aims to embed deep research into business apps, automate multi-agent processes, and deliver governable, auditable knowledge work at scale.
Key features:
- Web-scale automated research grounded with Bing Search.
- Programmable research agents callable from apps or workflows.
- Integration with Logic Apps, Azure Functions, and Foundry connectors.
- Enterprise-grade security, compliance, and transparency.
The Deep Research model, called o3-deep-research, handles multi-step pipelines: it scopes tasks via GPT-4 variants, grounds queries with Bing Search, analyzes and synthesizes info deeply, and outputs structured, fully auditable reports.
Developers can trigger research agents anywhere, enabling workflows like research → slide generation → email delivery.
Pricing:
- Input tokens: $10 per 1M
- Cached input: $2.50 per 1M
- Output tokens: $40 per 1M
Charges for Bing Search and GPT base clarify calls apply separately.
Deep Research is now in limited public preview for Azure AI Foundry Agent Service customers. Microsoft expects businesses across markets to use this for market analysis, intelligence, analytics, and compliance reporting.
"Deep Research in Foundry Agent Service is built for developers who want to move beyond the chat window…"
"Unlike packaged chat assistants, Deep Research… can evolve with your needs—ready for automation, extensibility, and integration with future internal data sources as we expand support."
Get started on Azure AI Foundry. This rollout pushes deep AI research from experiment to enterprise-grade service.
