Perplexity’s new Deep Research enables in-depth research for free

Perplexity has introduced its own Deep Research, its new deep reasoning model capable of independently conducting advanced research and expert-level analysis of complex topics. To accomplish this, this AI executes dozens of searches, analyzes hundreds of sources, and carries out a reasoning process on this information to create a comprehensive and highly useful summary.

For now, this new capability is available in the web version of Perplexity and will soon be rolled out on iOS, Android, and Mac. Additionally, it is noteworthy that it has been activated for all users.

Individuals using the free version of Perplexity can use Deep Research at no cost, although they will only have 5 queries per day. Meanwhile, those subscribed to the Pro plan will enjoy 500 queries per day.

Perplexity can generate detailed research reports

The model by Perplexity is not the first Deep Research model we have encountered. Earlier this month, OpenAI launched its own Deep Research as a reasoning agent integrated into ChatGPT. Like this AI, Perplexity’s has been designed to conduct deep research, which would take a human person hours, in a matter of minutes.

This makes it a perfect ally when conducting research or professional analysis across a multitude of fields. “We created Deep Research so that anyone can conduct expert-level analysis on complex topics. It is ideal for generating reports in areas such as finance, marketing, and technology, but it can also serve as a personal consultant in fields like health, product research, and travel planning,” explains Perplexity.

Furthermore, this AI has achieved high standards in various tests, surpassing the results obtained by models from other companies. For instance, Perplexity’s Deep Research achieved a 21.1% accuracy in Humanity’s Last Exam (a test with over 3,000 questions across more than 100 topics such as mathematics, sciences, history, or literature), ranking above prominent models like Gemini Thinking, o3-mini and DeepSeek R1.

In the benchmark SimpleQA (a database with thousands of questions that assess the accuracy and truthfulness of the information presented by a system), the model achieved an accuracy of 93.9%, far surpassing o1-mini, GPT-4o and Claude 3.5 Sonnet, among others.

Lastly, Perplexity emphasizes that its deep reasoning model “not only achieves high scores in industry tests but does so by completing most research tasks in under 3 minutes, and we continue to work to make it even faster”.

How Perplexity’s Deep Research works

To use Perplexity’s Deep Research, you simply need to access the AI, expand the model selector integrated into the search bar, and select this option. After that, enter your query into the tool.

The deep reasoning model of Perplexity will initiate a reasoning-based research process powered by its search and encoding capabilities. “Perplexity’s Deep Research iteratively searches, analyzes documents and decides the next steps, refining its research plan as it delves deeper into the topic. It is similar to how a human researches a new topic, adjusting their understanding throughout the process”.

Once it has analyzed all the sources, the AI will synthesize all the gathered information into a detailed report. This process will take between 2 and 4 minutes.

Finally, you will be able to export this report in PDF or document format, and you will also have the option to convert it into a Perplexity Page that you can share with your friends.

Photo: Perplexity