Google has enhanced its Gemini Advanced service by integrating the experimental 2.5 Pro model into its Deep Research feature. This upgrade aims to provide subscribers with more comprehensive and insightful research capabilities.
Initially, Deep Research was powered by the 2.0 Flash Thinking model and made available to all users. With the recent upgrade, paying subscribers now have access to the more advanced 2.5 Pro model. Currently, this feature is exclusive to subscribers, with free users not having access at this time.
Google reports that evaluations favored the reports generated by the 2.5 Pro-powered Deep Research over other leading providers by a margin greater than 2-to-1. The assessment focused on several criteria:
- Instruction Following: Adherence to user-specified instructions and constraints.
- Comprehensiveness: Breadth and range of information covered, addressing the full scope of user requests.
- Completeness: Depth and thoroughness of information on addressed topics.
- Writing Quality: Clarity, conciseness, logical organization, and overall readability of the report.
Overall, the preference was 69.9% for Gemini’s reports compared to 30.1% for those from OpenAI’s Deep Research. Users have noted significant improvements in analytical reasoning, information synthesis, and the generation of more insightful research reports.
To utilize the 2.5 Pro-powered Deep Research, subscribers can select the 2.5 Pro (experimental) option in the model picker on gemini.google.com. The “Deep Research” feature will then appear alongside “Canvas” in the prompt bar. This update is gradually rolling out to Android and iOS devices. Additionally, Google has recently reordered the available models in the picker for improved user experience.
This development underscores Google’s commitment to advancing AI-driven research tools, offering users enhanced capabilities for their professional and personal research needs.
