Google Introduces 'Deep Research Max' AI Agent Powered by Gemini 3.1 Pro

Google has announced a major upgrade to its AI research capabilities with the launch of two new autonomous agents: Deep Research and Deep Research Max. Both are powered by the Gemini 3.1 Pro model and represent a significant evolution from the preview version released in December 2025. Deep Research is optimized for speed and low latency, designed for interactive user interfaces. Deep Research Max, the flagship offering, is engineered for maximum comprehensiveness and high-quality analysis, making it ideal for asynchronous background workflows like in-depth market research or financial analysis.

A key advancement is the agents' ability to safely navigate both the public web and private, custom data repositories—including internal documents and specialized financial streams—via Model Context Protocol (MCP) support. This allows for secure analysis of restricted enterprise environments. Deep Research Max has demonstrated impressive benchmark scores, including 93.3% on Google's DeepSearchQA for search and question-answering, and 54.6% on the challenging Humanity's Last Exam (HLE) benchmark. Furthermore, it now integrates native chart and infographic generation (using Nano Banana image generation) and HTML page creation, enabling it to visualize complex datasets and produce professional-grade, fully cited reports with more sources and better nuance than its predecessor.