Google has made significant strides in autonomous AI by open sourcing DeepSearchQA and providing access to the Gemini Deep Research Agent, powered by the advanced Gemini 3 Pro model. This agent mimics human researchers, executing multi-step web searches, refining questions, and validating sources for more reliable insights compared to traditional AI systems. DeepSearchQA serves as an open-source benchmark, featuring 900 tasks across 17 subjects—such as climate and health—to evaluate AI’s complex reasoning capabilities. The approach allows for assessing logic and accuracy over extended inquiries, positioning DeepSearchQA as an industry-standard guide for AI development. Developers can leverage the Deep Research API for document parsing, structured reports, and enhanced source citations, with upcoming features like chart creation and improved Model Context Protocol support. Additionally, the newly launched Interactions API enables dynamic agent interactions and background processing, enhancing Google’s push towards fully autonomous AI agents that conduct in-depth research and decision-making.
Source link
