Perplexity Moves Deep Research Into Computer, Routing Research Subtasks Across 20+ Frontier Models For Reports, Decks, And Dashboards

Perplexity has moved Deep Research into Computer, its multi-model orchestration system. The improve improves accuracy, depth of study, and quotation high quality. Deep Research now breaks onerous questions into subtasks and routes them throughout 20+ frontier fashions. It returns work-ready experiences, decks, and dashboards, all inside Computer.

Deep Research in Computer

Deep Research is a mode that runs many searches, reads sources, and writes a cited report. The new model lives inside Perplexity Computer, which launched in late February 2026. Computer is a cloud system that coordinates as much as 20 AI fashions in a single workflow. It is model-agnostic, with Opus 4.6 as its core reasoning engine. Sub-agents deal with specialised work, reminiscent of Gemini for deep analysis duties.

Deep Research in Computer is constructed on two components: the Agent Search SDK and Search as Code. With one advanced query, it builds a analysis plan mechanically. It then finds main sources throughout lots of of web sites and cites each declare.

Search as Code: How It Works

The mannequin writes code that assembles the search itself. That code runs hundreds of retrieval steps in parallel, tailor-made to every query. The script runs in a sandbox and calls Perplexity’s Agentic Search SDK. The SDK exposes search primitives reminiscent of filtering, deduplication, and reranking. This differs from a set pipeline that runs the identical steps each time. Code-driven search lets the system department, examine, and refine because it learns.

Search as Code is rolling out by each Computer and the Agent API. So builders can attain the identical agentic search stack programmatically. Computer additionally reads your recordsdata alongside the reside internet. You can pull in a PDF or spreadsheet for inside context. It then cross-references that in opposition to census information, Statista, and different sources.

A Working Developer Example

Deep Research in Computer is a client characteristic for Perplexity Max customers. Developers attain the identical stack by the pay-as-you-go Agent API. The official SDK ships a deep-research preset, proven under.

# pip set up perplexityai
# export PERPLEXITY_API_KEY="your_api_key_here"
from perplexity import Perplexity

consumer = Perplexity()  # reads PERPLEXITY_API_KEY from the atmosphere

response = consumer.responses.create(
    preset="deep-research",  # pre-configured analysis setup; "pro-search" is one other
    enter="Compare the money stream and revenue margins of the biggest AI chip makers.",
)

print(response.output_text)  # aggregated report textual content from the run

The endpoint is POST https://api.perplexity.ai/v1/agent. It additionally accepts POST /v1/responses for OpenAI SDK compatibility.

Benchmark

Perplexity revealed before-and-after numbers evaluating legacy Deep Research with the Computer model. The positive aspects are largest on agentic searching, the place the system should navigate many pages.

BrowseComp checks an agent’s skill to seek out hard-to-locate info by searching. The bounce from 40.7% to 83.8% is the biggest acquire proven. Humanity’s Last Exam covers skilled questions throughout many tutorial topics. DeepSearchQA already sat excessive, so its acquire is smaller however optimistic.

Use Cases, With Examples

Perplexity ships starter duties that present the supposed scope.

• Finance: examine money stream and revenue margins of main AI chip firms over 5 years.
• Legal: map how US and European data-privacy legal guidelines differ into one comparability desk.
• Healthcare: synthesize clinical-trial proof on whether or not weight-loss medicine enhance coronary heart well being.
• Technology: benchmark main fashions on reasoning skill, price, and context size.

Each job ends in a deliverable. You can flip a report into a short, a deck, or a reside spreadsheet. Computer reads and writes contained in the file, not beside it. It exhibits a preview earlier than any change lands, which you approve or reject.

How It Picks Models

Computer routes every subtask to the mannequin greatest suited to it. A authorized reasoning mannequin handles contract assessment. A knowledge mannequin handles spreadsheet variance checks. A writing mannequin handles the ultimate draft. Premium information sources again the solutions, together with PitchBook and CB Insights. Legal information is presently in preview.

Strengths and Limitations

Strengths:

• Code-driven search runs hundreds of retrieval steps in parallel per query.
• Large measured positive aspects on agentic searching, led by the BrowseComp outcome.
• Reads inside recordsdata and the reside internet, citing each declare inline.
• Produces prepared deliverables: experiences, briefs, decks, dashboards, and reside spreadsheets.

Limitations:

• The benchmark numbers are first-party, so unbiased verification nonetheless issues.
• The in-Computer characteristic facilities on Perplexity Max, not a free tier.
• Premium-source protection varies, and authorized information stays in preview.
• Outputs nonetheless want human assessment, since “cited” doesn’t all the time imply right.

Key Takeaways

• Perplexity moved Deep Research into Computer, routing analysis subtasks throughout 20+ frontier fashions.
• “Search as Code” lets the mannequin write code that runs hundreds of retrieval steps in parallel.
• BrowseComp accuracy jumped from 40.7% to 83.8%; Humanity’s Last Exam rose 36.4% to 50.5%.
• It reads your recordsdata and the reside internet, citing each declare throughout experiences, decks, and dashboards.
• Developers can attain the identical agentic search stack by the pay-as-you-go Agent API.

Check out the Technical details. Also, be at liberty to comply with us on Twitter and don’t overlook to affix our 150k+ ML SubReddit and Subscribe to our Newsletter. Wait! are you on telegram? now you can join us on telegram as well.

Need to associate with us for selling your GitHub Repo OR Hugging Face Page OR Product Release OR Webinar and so on.? Connect with us

The put up Perplexity Moves Deep Research Into Computer, Routing Research Subtasks Across 20+ Frontier Models For Reports, Decks, And Dashboards appeared first on MarkTechPost.