Google's Gemini Push

Google has folded its newest research agent, Deep Research Max, into Gemini 3.1 Pro, extending the model family it introduced in preview on February 19 and updated again this week. (blog.google) A research agent is software that plans a question, searches for material, and writes a cited report; Google says Gemini Deep Research now does that with collaborative planning, chart support, file inputs, and connections to external tools through Model Context Protocol servers. (ai.google.dev 1) (ai.google.dev 2) Google’s pitch is that Gemini 3.1 Pro supplies the reasoning layer underneath. In its February 19 launch, Google said 3.1 Pro was available in preview in Vertex AI, Gemini Enterprise, Google AI Studio, Android Studio, Gemini CLI, and other developer surfaces. (cloud.google.com) (blog.google) The model is built for large, messy inputs rather than short prompts. Google’s model card says Gemini 3.1 Pro accepts text, images, audio, video, and code with a context window of up to 1 million tokens and up to 64,000 output tokens. (deepmind.google) (docs.cloud.google.com) That matters because “deep research” products are becoming a distinct category inside consumer chatbots and enterprise software. Google said this week that Deep Research has shifted from a summarization tool into a starting point for finance, life sciences, and market-research workflows that feed larger agent pipelines. (blog.google) Google is also tying the product to a broader agent-building stack. Its developer documentation says the Deep Research agent can return detailed cited reports, stream results to a client interface, and use File Search and external MCP servers to pull in company data or other connected sources. (ai.google.dev 1) (ai.google.dev 2) The company has been laying the groundwork for months. In December 2025, Google opened Gemini Deep Research to developers through its Interactions API, alongside a benchmark called DeepSearchQA for complex web-search tasks. (blog.google) Google’s own product pages frame 3.1 Pro as the company’s top reasoning model for complex tasks. DeepMind’s model page calls it Google’s “most intelligent model yet,” while the model card says it is the company’s most advanced model for complex tasks as of the card’s February 2026 publication date. (deepmind.google 1) (deepmind.google 2) Pricing and access show where Google thinks this goes next. Google’s Gemini 3 developer guide lists Gemini 3.1 Pro Preview at $2 per 1 million input tokens and $12 per 1 million output tokens for prompts under 200,000 tokens, with higher rates above that threshold. (ai.google.dev) The immediate story is not a single chatbot feature but a packaging move: Google is turning long-context reasoning, web research, tool use, and report generation into one Gemini 3.1 Pro-centered product line. (blog.google) (blog.google)