Google launches Deep Research Agent: Quickly master automated 160-search cycles to generate charts, touted as research-grade collaborative AI

📄Full Article· Automatically extracted by trafilaturaGemini 翻譯1402 words

Google AI has officially launched two autonomous research agents, Deep Research and Deep Research Max, built on Gemini 3.1 Pro. They can automatically plan research paths, execute up to 160 web searches, connect to internal enterprise databases (via MCP), and generate charts directly within reports. Moving from "chatting with AI" to "letting AI run the entire research process" marks an upgrade in Google's definition of AI agents. (Context: Google upgrades Gemini Deep Research Max: Integrates MCP for internal database access and native charts, enabling analyst-level due diligence.) (Background: Jensen Huang sends an all-hands email embracing OpenAI Codex: Over 10,000 NVIDIA employees are already on board, with GPT-5.5 running on GB200.) If you have used Gemini's standard chat function for research, you likely know its limits: ask a question, get an answer, run a few searches at most, and then you have to piece the conclusions together yourself. The Deep Research series works differently; it discusses the research plan with you first, and once you approve, it runs autonomously in the background for up to 60 minutes, delivering a complete report with charts upon completion. Google has launched two versions, Deep Research and Deep Research Max, which differ in speed and depth. How to choose between the two versions: Deep Research is the speed-optimized version, running approximately 80 searches per task, suitable for real-time interaction and quick queries. You input a question and get results within minutes. Deep Research Max is the comprehensiveness-optimized version, running approximately 160 searches per task, with a token consumption 3-4 times that of the standard version (approx. 900k input + 80k output), making it suitable for deep analysis tasks that can run overnight. It iteratively refines the report, which Google calls "extended test-time compute." Regarding costs, the standard version is approximately $1-3 per task, while the Max version is approximately $3-7. Billing is based on the underlying Gemini 3.1 Pro token rates: $2.00 per million input tokens and $12.00 per million output tokens. Three key features: MCP integration for enterprise databases: Through the Model Context Protocol (MCP), you can connect Deep Research to your own data sources. Google is collaborating with FactSet, S&P Global, and PitchBook to develop MCP integrations, allowing those performing financial due diligence to have the agent query corporate financial databases directly instead of just searching the public web. You can also disable web search entirely and run it solely on internal data. Native chart output: Reports can directly generate HTML charts or infographics, eliminating the need to manually paste data into Excel to create graphs. This serves as an intermediate product between a plain text report and a visual analysis file, as you can still perform secondary edits. Collaborative research planning: The agent does not run immediately; it first generates a research plan for you to review, modify, and confirm before execution. If you are dissatisfied with the results, you can continue to follow up or adjust the direction through multi-turn conversations. Both agents are currently in Public Preview, available via the paid tier of the Gemini API. If you want to try them out, you can operate them directly on Google AI Studio. Support is already available in the Python SDK and JavaScript SDK; those with free credits should not miss out.

Data Status✓ Full text extractedRead Original (動區 BlockTempo)

🔍Historical Similar Events· Keyword + Asset Matching6 items

2026-04-23

PrimePiper: A prime broker for AI agent trading, enabling AI agents to trade securely in global markets.

Similarity 120%關鍵字 agent同分類 zh

2026-04-23

Introduction to Claude Code's new feature /ultrareview: Cloud-based multi-agent deep code review, now free for a limited time for Pro and Max users

Similarity 120%關鍵字 agent同分類 zh

2026-04-23

Beating GPT-5.4 at 2 cents per query: Perplexity reveals post-training recipe for search agents

Similarity 120%關鍵字 agent同分類 zh

2026-04-23