Google DeepMind announced two significant evolutions of its autonomous research agent: Deep Research and Deep Research Max. Both are built on Gemini 3.1 Pro, and together they represent a meaningful shift in what AI-assisted research can actually do at a professional level.
If you’ve been following the AI space, you’ll know that “research agents” have been a hot topic. But most of them still feel like glorified summarizers. Google is making a credible case that it’s moving past that.
From Summarizer to Research Engine
When Google first released the Gemini Deep Research agent to developers back in December 2025 via the Interactions API, it was already impressive. But the team describes the new version as a transformation from “a sophisticated summarization engine into a foundation for enterprise workflows.” That’s not just marketing language. The new agents are designed to support real-world use cases in finance, life sciences, and market research, producing fully cited, professional-grade analyses from a single API call. The key upgrade? The ability to blend the open web with proprietary data streams, which changes the game significantly for enterprise users.
Two Agents, Two Use Cases
Google is now offering two distinct configurations. Deep Research is the speed-optimized option, built for interactive user surfaces where low latency matters. Think of it as the version you’d embed directly into a product that a user is actively engaging with. Deep Research Max is the heavy-duty option, using extended test-time compute to iteratively reason, search, and refine its output. Google specifically positions it for asynchronous, background workflows, like a nightly cron job that generates exhaustive due diligence reports for an analyst team by morning.
MCP Support, Native Charts, and Multimodal Inputs
Three new capabilities stand out. First, Model Context Protocol (MCP) support is arguably the most significant addition. Deep Research can now connect to custom data sources and specialized professional data providers (think financial data, market intelligence platforms) securely via MCP. This turns it from a web searcher into an autonomous agent capable of navigating any specialized data repository you point it at. Google is already collaborating with FactSet, S&P Global, and PitchBook on their MCP server designs.
Second, native charts and infographics are new for the Gemini API. The agent now generates high-quality charts and infographics inline with HTML, dynamically visualizing complex data sets within the report itself. That’s a meaningful step toward presentation-ready outputs.
Third, multimodal research grounding means you can feed the agent a mix of PDFs, CSVs, images, audio, and video as input context. Combined with Google Search, URL context, code execution, and file search running simultaneously, the tool is starting to look like a genuine research analyst rather than a chatbot with a search button.
Collaborative Planning and Real-Time Streaming
Two other additions are worth noting for developers and product builders. Collaborative planning lets you review and refine the agent’s research plan before it begins execution, giving you granular control over scope. This is critical in regulated industries where you need to know exactly what the agent is and isn’t looking at. Real-time streaming, meanwhile, lets you track the agent’s intermediate reasoning steps live, with thought summaries and outputs arriving as they’re generated, which is a much better experience than waiting for a finished report to land.
The Enterprise Bet
What’s clear from this announcement is that Google is making a serious enterprise play. The focus on finance and life sciences, the partnerships with FactSet, S&P Global, and PitchBook, and the emphasis on factuality and source diversity all point to a product team working in regulated, high-stakes environments. Google also notes that Deep Research already powers research capabilities inside the Gemini App, NotebookLM, Google Search, and Google Finance, which is a meaningful signal about the maturity of the underlying infrastructure.
Deep Research and Deep Research Max are available today in public preview via paid tiers in the Gemini API, with Google Cloud availability for startups and enterprises coming soon. If you’re building research-heavy products or workflows, this is worth a close look.
Originally posted on my Substack
Top comments (0)