Google has rolled out a major upgrade to its Gemini Deep Research agent.
At the core of the upgraded agent is Gemini 3 Pro, Google’s most factual model yet.
Google is also open-sourcing DeepSearchQA benchmark.
Google has rolled out a major upgrade to its Gemini Deep Research agent, giving developers access to a far more powerful system for long-form research, analysis and information gathering. Interestingly, Google’s announcement arrived on the same day OpenAI launched GPT-5.2. Instead of returning quick answers, the agent plans its work carefully: it creates search queries, reads through results, identifies what it still doesn’t know, and searches again. This process helps it gather deeper, more accurate information from across the web.
SurveyAt the core of the upgraded agent is Gemini 3 Pro, Google’s most factual model yet. It has been trained specifically to reduce hallucinations and to produce clearer, more reliable reports. According to the tech giant, the upgraded Gemini Deep Research agent achieves “state-of-the-art results on Humanity’s Last Exam (HLE) and DeepSearchQA, and is our best on BrowseComp.” “Deep Research is now more useful and intelligent than ever, and will soon be available in Google Search, NotebookLM, Google Finance and upgraded in the Gemini App.”
Also read: OpenAI brings GPT 5.2 to take on Gemini 3 Pro, Sam Altman says its most capable model yet
Developers can use the Deep Research agent to analyse uploaded documents, combine them with web findings, and generate structured reports. It also supports custom formatting, which means you can control the output via prompting.
Google says future updates will bring built-in chart generation, better connections to custom data sources through MCP, and availability through Vertex AI for enterprise use.
Also read: OpenAI’s ChatGPT can now edit your images using Adobe Photoshop: Here is how
The tech giant is also open-sourcing DeepSearchQA, a benchmark built to test how well research agents handle long, multi-step tasks. The benchmark includes 900 carefully designed tasks across 17 fields. Unlike simple fact-checking datasets, DeepSearchQA “measures comprehensiveness, requiring agents to generate exhaustive answer sets. This assesses both research precision and retrieval recall,” Google explains.
Also read: US attorneys general warn OpenAI, Google and other AI giants to fix delusional chatbot outputs
Follow Us
Ayushi Jain
Tech news writer by day, BGMI player by night. Combining my passion for tech and gaming to bring you the latest in both worlds. View Full Profile
