Perplexity launched its Computer platform in late February 2026. The idea was simple to describe but hard to build: instead of one AI model answering your question, a coordinating layer picks the right specialist model for each part of the job and runs them all at once. Deep Research is now fully integrated into that system.
The practical meaning for a busy professional: you type one question and get back a finished document. Not a summary you still have to turn into a report. An actual report, or a deck, or a spreadsheet. The system reads your uploaded files alongside live web sources, pulls data from paid databases like PitchBook and Statista, and attaches a citation to every claim.
The performance numbers Perplexity published are worth paying attention to. On a web-browsing test designed to measure how well an AI tracks down hard-to-locate information, the score jumped from 40.7% to 83.8%. On a test covering expert-level questions across many academic fields, the score moved from 36.4% to 50.5%. Both tests are created by independent AI safety and research organizations, not Perplexity itself.
What changed under the surface is that the system no longer runs the same fixed search process for every question. Instead, it writes its own search script tailored to each question, then runs thousands of individual information-retrieval steps in parallel across those 20+ models. One model handles legal reasoning. Another handles data analysis. A writing model produces the final draft. The system shows you a preview before making any changes to a file, so you stay in control.
Premium data sources are included: PitchBook for company and investor data, CB Insights for market analysis, Statista for industry statistics. Legal data access is in preview. For anyone who has paid separately for even one of those subscriptions, the comparison is immediate: a single Statista license alone costs far more than a Perplexity Pro subscription.
The access question is straightforward. The full Computer system, including Deep Research in its new form, requires the Max plan at $200 per month. The Pro plan at $20 per month gives research mode but without the multi-model depth. For teams, Enterprise Pro starts at $40 per seat per month and adds privacy controls and shared workspaces. Enterprise Max, at $325 per seat per month, removes usage caps.
Perplexity also added Computer inside Microsoft 365 apps in May, meaning the same research workflow now runs directly inside Word, Excel, PowerPoint, Outlook, and Teams. That matters for organizations where work lives inside Office tools rather than standalone platforms.
A hybrid local-cloud system is coming in July 2026. It will automatically decide, task by task, whether to process something on your local device or send it to the cloud, with the stated goal of keeping sensitive data on-device unless cloud processing is explicitly needed.
The broader picture is that Perplexity is no longer just a smarter search engine. It is positioning itself as the layer that handles research and synthesis end-to-end, across whatever AI models happen to be best at any given moment. Whether the $200 monthly price makes sense depends entirely on how much of your team's time currently goes into pulling, organizing, and citing information from multiple sources. For research-heavy roles, the calculation is getting harder to ignore.