Google launches Deep Research Max for AI agents

Google is making two autonomous research agents available to developers: Deep Research and Deep Research Max. Both can search for information, cross-check sources, and generate cited reports, but they are designed for different workflows.

The update integrates Gemini 3.1 Pro and takes the system beyond automatic summaries. With a single API call, a company can combine information from the internet with its own data to create analysis for finance, life sciences, market research, and other industries.

Two agents for two types of work

Deep Research replaces the preview version introduced in December. It is optimized for faster responses, lower costs, and higher quality, making it a better fit for interactive tools where users expect results quickly.

Deep Research Max prioritizes depth. It uses more computing capacity during execution to investigate, reason, review its findings, and iteratively refine the final report.

Its goal is not to respond in seconds, but to complete processes that can run in the background. For example, a company could schedule a due diligence analysis every night so an analyst team finds it ready in the morning.

Google says Max consults more sources than the December version and detects nuances that could previously go unnoticed. It is also designed to compare contradictory evidence and give more weight to specialized sources, such as SEC filings or peer-reviewed academic papers.

It can also work with your data

The agent is no longer limited to the web. It can query uploaded files, connected document stores, and external services through MCP, a standard that connects AI models to custom tools and databases.

That makes it possible to use professional data that is normally restricted. Google cites integrations with financial and market data providers such as FactSet, S&P Global, and PitchBook. Their MCP servers are being developed so their customers can add them to these workflows.

Developers can combine Deep Research with several Gemini API tools:

• Google Search to query the web.
• Remote MCP servers to access proprietary or specialized data.
• URL Context, Code Execution, and File Search.
• PDF, CSV, image, audio, and video files as context.

It is also possible to disable web access completely and base the research solely on internal information. This is especially useful for companies handling confidential data or needing precise control over the sources used.

Reports with charts and more control

Deep Research can generate charts and infographics within the report itself, using HTML or Nano Banana. The idea is to turn complex datasets into visualizations that can be understood and shared without manually preparing the document afterward.

You do not have to accept the research plan at face value, either. You can review the strategy proposed by the agent, adjust its scope, and then let it begin searching. The system also provides live summaries of its intermediate steps and lets you receive text and images as they are generated.

For you, the most important change is that these agents can move from searching for information to carrying out an entire part of the analysis process. Their reports still need human review, especially for financial, scientific, or regulatory decisions, where a misinterpreted source can have real consequences.

Deep Research and Deep Research Max are available starting today in public preview, as part of paid Gemini API plans and through the Interactions API. Google also plans to bring them to startups and enterprises through Google Cloud soon.

What to watch now is whether Max's greater depth justifies the additional time and cost compared with the faster version. The direction is clear: AI agents are starting to move beyond simple search and become tools that gather context, query private systems, and deliver analysis ready to be integrated into professional workflows.