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

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 upgrade improves accuracy, depth of analysis, and citation quality. Deep Research now breaks hard questions into subtasks and routes them across 20+ frontier models. It returns work-ready reports, decks, and dashboards, all inside Computer.

Deep Research is a mode that runs many searches, reads sources, and writes a cited report. The new version lives inside Perplexity Computer, which launched in late February 2026. Computer is a cloud system that coordinates up to 20 AI models in one workflow. It is model-agnostic, with Opus 4.6 as its core reasoning engine. Sub-agents handle specialized work, such as Gemini for deep research tasks.

Deep Research in Computer is built on two parts: the Agent Search SDK and Search as Code. With one complex question, it builds a research plan automatically. It then finds primary sources across hundreds of sites and cites every claim.

The model writes code that assembles the search itself. That code runs thousands of retrieval steps in parallel, tailored to each question. The script runs in a sandbox and calls Perplexity’s Agentic Search SDK. The SDK exposes search primitives such as filtering, deduplication, and reranking. This differs from a fixed pipeline that runs the same steps every time. Code-driven search lets the system branch, compare, and refine as it learns.

Search as Code is rolling out through both Computer and the Agent API. So developers can reach the same agentic search stack programmatically. Computer also reads your files alongside the live web. You can pull in a PDF or spreadsheet for internal context. It then cross-references that against census data, Statista, and other sources.

Deep Research in Computer is a consumer feature for Perplexity Max users. Developers reach the same stack through the pay-as-you-go Agent API. The official SDK ships a deep-research preset, shown below.

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

Perplexity published before-and-after numbers comparing legacy Deep Research with the Computer version. The gains are largest on agentic browsing, where the system must navigate many pages.

BrowseComp tests an agent’s ability to find hard-to-locate information through browsing. The jump from 40.7% to 83.8% is the largest gain shown. Humanity’s Last Exam covers expert questions across many academic subjects. DeepSearchQA already sat high, so its gain is smaller but positive.