Perplexity enhances search with o1-powered reasoning and new tools

Perplexity recently shipped a couple of changes, with the biggest one still likely in the process of rolling out. This update is focused on integrating o1-powered reasoning into their cross-search workflow. In the past, reasoning only worked in a separate focus mode during beta, with a limit of 10 queries per day, and didn’t have access to web search. Now, it works differently. While using the normal web search, for some queries, reasoning is automatically triggered, displaying a new UI with a progress bar.

Users have the option to skip the advanced reasoning if they wish. This feature is marked with a beta label, and for complex queries, it performs impressively well. For instance, users have asked Perplexity to search for OpenAI co-founders, track when they left the company, and highlight where they are working now. The new reasoning feature handled this task seamlessly, even plotting the results into a table. The logic behind when and why reasoning is triggered seems to be hidden and may be rate-limited, but there hasn’t been an official announcement yet, just subtle hints.

On a smaller note, Perplexity is adding support for 8 new languages. Six of these are European languages, in addition to the two previously discovered.

There’s also a new toggle in the settings that lets users turn off auto-suggestions.

Furthermore, Perplexity is working on a specific focus mode for universities. A recent promotional campaign targeted students, and it seems this new feature will allow students to search information across their respective .edu domains. So far, only the toggle for turning off search suggestions is available to all users, and the reasoning feature is being rolled out gradually. The new language support and the university focus mode are still in development.

Regarding reasoning, it appears that the overall results are generated using the 4o model. However, o1 reasoning is likely powering the cross-search functionality, allowing Perplexity to break down queries into tasks and plan what needs to be done. For the individual queries within the process, the regular 4o model is used. These changes make Perplexity Pro search one of the most powerful search tools in the world right now.