Last year, Google launched a new feature in Gemini called"deepdegree research"new features. Deep Research is an AI research assistant that can delve into complex topics and create reports with relevant citations. Along the way, it can browse up to hundreds of websites, reflect on its findings, and create reports that can also be turned into podcast-style conversations.

Last month, Google announced that the GeminiDeepResearch feature now uses the Gemini2.0 FlashThinkingExperimental model, which improves planning, search, reasoning, analysis, and reporting. The improved model provides users with higher quality multi-page reports.

Today, Google announced that deep research capabilities are now powered by its flagship cutting-edge model Gemini 2.5 Pro Experimental. GeminiAdvanced users can now use this feature on the web, Android and iOS. With powerful new models, deep research will enable better planning, search, reasoning, analysis, and reporting.

According to Google's own testing, human raters preferred reports generated by GeminiDeepResearch powered by Gemini2.5Pro to OpenAI's DeepResearch reports by more than 2 to 1. Users also noticed that this improved DeepResearch now offers significant improvements in analytical reasoning and information synthesis, and produces more insightful research reports.

If you are a GeminiAdvanced user, you can select the Gemini2.5Pro (experimental) model from the model selector drop-down menu and select "In-depth Study" in the prompt bar. This upgrade reflects Google's continuous efforts to enhance the functionality of GeminiAdvanced and provide users with more powerful tools to complete complex research tasks. You can also try GeminiDeepResearch and Gemini2.0 Flash models for free on desktop and mobile devices in 150 countries.