AI agent network connecting deep-research tasks across 20+ models via Gemini, illustrating Perplexity’s multi-model collabora

Editorial illustration for Perplexity routes deep‑research subtasks across 20+ models using Gemini agent

Perplexity routes deep‑research subtasks across 20+...

Perplexity routes deep‑research subtasks across 20+ models using Gemini agent

By AI Daily Post Edited by Brian Petersen, Editor-in-Chief

June 12, 2026 • 2 min read

Perplexity has shifted its Deep Research capability into Computer, the company’s new multi‑model orchestration platform that debuted in late February 2026. Why does that matter? Because the upgrade doesn’t just tack another search layer onto an existing pipeline—it breaks tough questions into subtasks and dispatches them across more than 20 frontier models, then stitches the results into a ready‑to‑use report, deck, or dashboard.

While the tech is impressive, the real change lies in the flexibility of the workflow. Opus 4.6 serves as the core reasoning engine, but the system is model‑agnostic; the SDK exposes primitives like filtering, deduplication and reranking, letting the process branch, compare and refine as it runs. Here’s the thing: “Search as Code” rolls out through both Computer and the Agent API, so developers can tap the same stack programmatically.

The platform also reads user‑provided PDFs or spreadsheets, cross‑referencing them against census data, Statista and other live sources. Deep Research in Computer is a consumer feature for Perplexity Max users, while the pay‑as‑you‑go Agent API offers the same stack to developers via a deep‑research preset.

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.

Search as Code: How It Works

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.

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

Why this matters

We see Perplexity’s shift of Deep Research into its Computer platform as a concrete step toward more modular AI workflows. By slicing a single query into subtasks and dispatching them to over twenty frontier models, the system claims higher accuracy, deeper analysis, and better citation quality. The inclusion of sub‑agents—Gemini handling deep‑research tasks, for example—suggests a move away from monolithic prompting toward specialized components.

For developers, the Agent Search SDK and Search as Code could lower the barrier to building custom research pipelines that automatically generate reports, decks, and dashboards. Founders may appreciate the promise of work‑ready outputs without hand‑crafting prompts, yet we remain uncertain whether the orchestration overhead will offset the gains in practice. Researchers will likely test the citation improvements, but the article does not reveal how primary sources are verified beyond the model layer.

In short, the upgrade illustrates a pragmatic integration of multiple models, but its real‑world impact on productivity and reliability is still to be measured.