Gemini 2.5 Pro Just Became Way More Useful For Heavy Research Tasks

gemini-2.5-pro-just-became-way-more-useful-for-heavy-research-tasks
Gemini 2.5 Pro Just Became Way More Useful For Heavy Research Tasks

4

Sign in to your Android Police account

Google Gemini

Summary

  • Google has added Deep Research support to the Gemini 2.5 Pro (experimental) AI model.
  • Deep Research builds a plan, scans sources, and delivers a tailored report on your topic.
  • Gemini 2.5 Pro outperformed OpenAI’s model in writing, accuracy, and overall quality.

Google has been upgrading Gemini at an impressively fast pace. While Gemini 2.0 launched recently in late December 2024, the new Gemini 2.5 Pro is already available. Initially limited to paid users, Google was quick to roll out Gemini 2.5 Pro (experimental) to free users. And now, the model is getting a major upgrade: support for Deep Research.

The Gemini app open on a tablet in front of a stylized background

Related

As first reported by 9to5Google, Gemini Advanced users can now access the Deep Research feature using the Gemini 2.5 Pro (experimental) model. Google had previously rolled out Deep Research for all users last month, but that version relied on the Gemini 2.0 Flash Thinking model. With this update, Gemini users can now use Google’s most intelligent model yet for deep research tasks.

If you’re unfamiliar, Deep Research works exactly as it sounds. You give Gemini a topic, and it builds a research plan, scans hundreds of websites, and generates a detailed report tailored to your needs.

See also  More Android Phones Can Now Start Using Google's Version Of Apple Handoff

Deep Research just landed on Gemini 2.5 Pro (experimental)

Gemini 2.5 Pro with Deep Research preference comparison with OpenAI's Deep Research model

Source: Google

Now with Gemini 2.5 Pro powering this feature — even though it is in the experimental stage — Google claims the reports are more accurate and insightful than ever. According to Google, “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.”

In comparison with OpenAI’s deep research model, Google claims Gemini 2.5 Pro (experimental) was preferred 69.9% of the time overall. It also outperformed OpenAI in instruction following (60.6% vs. 39.4%), comprehensiveness (76.9% vs. 23.1%), completeness (73.3% vs. 26.7%), and writing quality (58.2% vs. 41.8%).

You can access Deep Research with Gemini 2.5 Pro (experimental) by visiting gemini.google.com and selecting the Deep Research option. It’s also available in the Gemini app on Android and iOS. However, the feature is currently limited to paid users — though, knowing Google, it may roll out to free users eventually.