Google Unveils Gemini Deep Research Agent as OpenAI Launches GPT‑5.2

Google Expands Gemini Deep Research Capabilities

Google revealed a reimagined version of its Gemini Deep Research agent, leveraging the Gemini 3 Pro foundation model. The updated agent is designed not only to generate research reports but also to allow developers to embed Google’s advanced research functions directly into their own applications through a new Interactions API. This API gives developers greater control as the industry moves toward more autonomous, agent‑driven AI experiences.

The Gemini Deep Research tool can synthesize extensive information and manage large context dumps within a single prompt. Google cites use cases ranging from due‑diligence investigations to drug‑toxicity safety studies, emphasizing the agent’s ability to handle complex, multi‑step reasoning tasks while minimizing hallucinations—a known challenge for large language models.

Integration Across Google Services

Google plans to weave the Deep Research agent into several of its flagship services, including Google Search, Google Finance, the Gemini app, and the popular NotebookLM platform. By embedding the agent into these products, Google aims to shift the user experience from manually searching the web to relying on AI agents that can autonomously retrieve and analyze information.

New Benchmarks to Measure Agent Performance

To substantiate its performance claims, Google introduced a benchmark called DeepSearchQA, targeting complex, multi‑step information‑seeking tasks. The benchmark is open‑sourced, allowing broader community evaluation. Google also evaluated the agent on existing benchmarks such as Humanity’s Last Exam—a general‑knowledge test with niche tasks—and BrowserComp, which assesses browser‑based agentic capabilities. In these evaluations, the Gemini Deep Research agent outperformed competing systems on its own benchmark and Humanity’s Last Exam, while OpenAI’s ChatGPT 5 Pro placed a close second and slightly edged Google on BrowserComp.

OpenAI’s Countermove: GPT‑5.2 (Garlic)

On the same day as Google’s announcements, OpenAI launched GPT‑5.2, codenamed Garlic. OpenAI claims the new model surpasses rivals, including Google’s offerings, across a suite of standard benchmarks and its own internal tests. The timing of the releases suggests a competitive race to dominate the emerging market for autonomous AI agents.

Industry Context and Future Outlook

The simultaneous announcements underscore a rapidly intensifying competition between major AI players. Both companies are emphasizing reduced hallucination rates, deeper reasoning abilities, and broader integration into consumer‑facing products. As AI agents become more capable, the balance between user control and autonomous decision‑making will shape the next phase of AI adoption.