Deep Research Powered By Gemini 2.5 Pro (Experimental) Now Available

Gemini Advanced subscribers can now perform Deep Research with Google’s most intelligent 2.5 Pro (experimental) model.

When Google made Deep Research available to all users last month, it was powered by 2.0 Flash Thinking (experimental). Paying customers can now use it with 2.5 Pro. Free users don’t have access to this today.

Google says “raters preferred the reports generated by Gemini Deep Research powered by 2.5 Pro over other leading deep research providers by more than a 2-to-1 margin.” Specifically, Google compared it against OpenAI Deep Research across:

Instruction following evaluates response’s fidelity to user specified instructions and constraints
Comprehensiveness measures breadth and range of information covered in response, addressing the scope of user request
Completeness measures the depth and thoroughness of information for topics addressed in the report
Writing quality evaluates clarity, conciseness, logical organization and overall readability of the report

The “Overall” is 69.9% vs. 30.1% with research questions “spanning professional (e.g., academic research, industry research) and personal (e.g., travel, shopping) use cases.”

Users report a “noticeable improvement in analytical reasoning, information synthesis and generating even more insightful research reports.”

After selecting 2.5 Pro (experimental) in the model picker on gemini.google.com, “Deep Research” now appears alongside “Canvas” in the prompt bar. This is still rolling out on Android and iOS. (Google also made a small tweak to the picker yesterday that reordered the available models.)