Google launched its top AI research agent – on the same day Openai released GPT-5.2

Google released on Thursday “reimagined” research agents from Gemini Deep Research based Foundation Modification Foundation-of-The-ArtGEMINI 3 PRO.

These new agents aren’t just designed to produce research reports – although that can still be done. It now allows developers to display Google’s sata-model research capabilities in their own applications. These capabilities are implemented through Google’s new Fire interactionwhich is designed to give sevs more control in the coming AII era.

The New Gemini Research Tool is an agent equipped to synthesize mountains of information and handle large contexts in directions. Google says it is used by customers for its initial tasks due to toxic security research.

Google also said it will integrate these new research agents into services, including Google Finance, the popular Gemini app. This is another step to prepare for a world where there is nothing more, AI agents.

The Tech giant says that the internal research of the status of Gemini 3 Pro is the “fantiest” model trained to minimize hallucinations during complex tasks.

Ai hallucinations – where LLM just makes things up – is an important problem for long tasks, where many autonomous tasks are created in minutes, hours, or longer. Another choice that LLM has to make, the greater is the choice that will not be forgotten by all.

To prove its claim of progress, Google also created other benchmarks (like the world needs one more). The sign of the new book can not be called Deepsearchqa, and its purpose is to test the agent in complex tasks, various steps of information. Google revealed this benchmark.

It also attempts deep research on the human final exam, which is interesting, an independent benchmark of general knowledge filled with acceptable tasks; And Browser, the benchmark for browser agent tasks.

As you might expect, Google’s new agent gives the competition its own benchmark, and humanity. However, Chatgpt 5 Pro OPENII is quite a bit of a surprise and a bit of a surprise to Google in BrowserComp.

But compared to the benchmark almost the growth of Google published them. Because on the same day, Openai released the much anticipated GPT 5.2 – garlic. Openai says that its latest model is that of competitors – especially Google – in a suite of typical benchmarks, including Homegrown Openai.

Perhaps one of the most interesting parts of this announcement is the timing. Knowing that the world is waiting for the onion, Google released some AI news of its own.