In the field of artificial intelligence, Google recently launched a "reimagined" Gemini deep research assistant, an update that connects developers with its most advanced autonomous research capabilities. The new system is built on Gemini 3 Pro and aims to transform deep research from a specialized report-writing assistant into an autonomous research agent capable of long-term reasoning and complex analysis.
The new system has the ability to analyze and condense large volumes of complex data, providing more detailed and accurate reports compared to previous versions. Google stated in a blog post that this updated system also reduces the occurrence of "hallucinations," which is crucial for long-running agents. Google emphasized: "The agent optimizes the ability to perform tasks involving long-term context collection and synthesis." Through large-scale multi-step reinforcement learning, the agent can autonomously navigate complex information environments with high accuracy.
During research, Deep Research is able to gradually plan its investigation strategy — it formulates queries, reads results, identifies knowledge gaps, and then searches again. At the same time, Google also introduced a new tool that allows developers to directly embed Google's research capabilities into their own applications. This new interface is designed to facilitate interaction between different agents, giving developers greater control within an increasingly powerful agent AI environment.
Google stated that early users have already applied Gemini Deep Research in areas such as financial services, biotechnology, market research, and drug toxicity safety studies, where precision and traceability are critical. Additionally, Google plans to integrate the newly released deep research agent into its various services, including Google Search, Google Finance, the Gemini app, and NotebookLM.
In addition to the release of the new assistant, Google also launched a new complex web search benchmark — DeepSearchQA, a framework specifically designed to evaluate agents' performance in "complex multi-step information retrieval tasks." Through this benchmark, Google found that when agents have more time for search and reasoning steps, their performance significantly improves, and they stated that they will continue to develop this further in future versions.
On the same day Google released this new system, OpenAI also launched its "most powerful" GPT-5.2 version, further intensifying the competition between the two companies in defining the next generation of agent AI systems.
Key Points:
🌟 Google launches a new version of Gemini Deep Research Assistant, enhancing autonomous research capabilities.
📊 The new system can analyze large volumes of complex data, reduce hallucinations, and provide more accurate reports.
🔍 On the same day, OpenAI releases GPT-5.2, marking a new phase in AI competition.
