Google has introduced Deep Research and Deep Research Max, powered by Gemini 3.1 Pro, marking a step change in its autonomous research agents. The update builds on the Gemini Deep Research preview released in December via the Interactions API, expanding it into a system capable of generating structured, fully cited reports.
Deep Research combines open web and proprietary data to support long-horizon workflows through a single API call, and serves as an entry point for agent-based pipelines focused on in-depth context gathering across domains such as finance, life sciences, and market research.
Deep Research Max
Deep Research Max is designed for comprehensive analysis and high-quality synthesis. It uses extended test-time compute to iteratively search, reason, and refine outputs, making it suitable for asynchronous workflows such as scheduled or overnight report generation.
It consults a broader range of sources, identifies nuanced insights, and evaluates conflicting information before producing structured outputs. The system draws from sources such as regulatory filings and peer-reviewed journals, presenting findings in formats suitable for analytical and stakeholder use.
The standard Deep Research agent replaces the preview version and is optimized for lower latency and cost, making it suitable for interactive applications.
Key features
- Unified research execution: Enables end-to-end workflows through a single API call, combining open web and proprietary data sources for fully cited outputs
- MCP-based data integration: Connects securely to external systems with flexible tool definitions and supports web, remote servers, file uploads, and connected storage with selective source usage
- Integrated visualization: Generates charts and infographics within reports using formats such as HTML and Nano Banana
- Controlled research workflow: Allows plan review and refinement with granular control, supports tools such as search and code execution, and enables disabling web access for custom-only research
- Multimodal and real-time processing: Accepts inputs such as PDFs, CSVs, images, audio, and video, while streaming intermediate summaries and outputs during execution
- Advanced source analysis: Expands coverage, compares sources, and evaluates conflicting information to improve consistency and depth
Deep Research supports automated workflows such as scheduled and large-scale research tasks, including overnight report generation. Google is working with organizations such as FactSet, S&P Global, and PitchBook on MCP server designs to enable integration of financial data platforms into these workflows.
The system runs on the same infrastructure used in Gemini App, NotebookLM, Google Search, and Google Finance, allowing it to scale across different research environments.
Availability
Deep Research and Deep Research Max are available in public preview through paid tiers of the Gemini API via the Interactions API. Google has stated that the features will also be made available to startups and enterprises through Google Cloud. Developers can begin using the agents through the available API documentation.