Deep Research and Deep Research Max: Google's autonomous research agents with MCP
In one sentence Google ships two research agents on the Gemini API: Deep Research (fast) and Deep Research Max (deep + slow, 93.3% on DeepSearchQA). MCP support for private data, native visualizations via Nano Banana 2.
"Deep research" is the pattern where you tell an AI "do a thorough analysis on X" and it goes off to read dozens of sources, cross-check, synthesize, and hand you a report. OpenAI and Anthropic have offered it since late 2024. Google ships its version on the Gemini API on April 21, 2026, and makes a move: two agents, your choice.
- Deep Research: fast, low-latency, low-cost. For when you want an answer in a few minutes.
- Deep Research Max: slow, luxurious, "spends compute". It iterates, searches again, reasons again. Designed for reports where quality matters more than time.
The novelty vs competitors is that they use MCP (the Model Context Protocol) to connect to your private data: your Drive, your CRM, the company database. Not just the web. And they generate charts and infographics inside the response (using Nano Banana 2 for images).
Stated benchmark: Deep Research Max scores 93.3% on DeepSearchQA (up from 66.1% in December). Available in public preview on paid Gemini API tiers.
Companies
Google, Google DeepMind
Tools
Deep Research, Deep Research Max, Gemini 3.1 Pro
Tags
Sources