Editorial illustration for Perplexity routes deep‑research subtasks across 20+ models using Gemini agent
Perplexity routes deep‑research subtasks across 20+...
Most AI search is still just a fancy text predictor. Perplexity decided to build a factory instead. Its new system breaks a single query into pieces and farms them out to over twenty different AI models, all working at once.
Gemini does the heavy research lifting. Other models pitch in with what they're good at. It runs on two things: an Agent Search SDK and something called Search as Code.
You ask a complicated question. The AI then writes a computer program whose sole job is to hunt down the answer. That script fires off thousands of tailored web searches in parallel, running in a safe sandbox.
It pulls primary sources from hundreds of sites. It cites everything. The output is a researched report, a presentation deck, a data dashboard—assembled by a swarm of specialized sub-agents.
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.
The goal is to make investigation automatic and auditable. It replaces a researcher's manual slog with a brief to a machine foreman. That foreman dispatches a whole team.
Each claim gets a footnote. Every source is logged. This is less about answering a question and more about building a documented case.
The model orchestra is the point. One AI isn't smart enough, so they use twenty. The search bar is becoming a project manager.
Related Reading
Browse all Google news →
Research & Benchmarks Google's FACTS benchmark shows 70% factuality ceiling across four tests
Research & Benchmarks Databricks finds multi-step agents beat single-turn RAG by 21% to 38% on STaRK
Research & Benchmarks Nvidia's DLSS 4.5 beta adds 6x Multi Frame Generation for RTX 50 GPUs
Related Topic NVIDIA and Google Cloud let developers scale AI from prototype to production
Market Trends
Related Topic Google tests visual 'magazine-style' UI for Gemini 3 Pro users
LLMs & Generative AI
From Archives Visual model exploits similarity of 打, 拍, 拉; text model starts from embeddings
From Archives New arXiv Paper Introduces Strategic Decision Support for AI Agents
From Archives Google sues Chinese Outsider Enterprise for Gemini-driven phishing on Telegram
LLMs & Generative AI
From Archives Gemini Omni adds AI video generation, using compute limits based on complexity and size
LLMs & Generative AI
Common Questions Answered
How does Perplexity's new system differ from traditional AI search engines?
Instead of using a single AI model to generate answers like a text predictor, Perplexity breaks down complex queries into subtasks and distributes them across over twenty different AI models working simultaneously. This approach treats search like a factory with specialized workers rather than a single generalist, allowing each model to contribute its specific strengths to solve different parts of the research problem.
What role does Gemini play in Perplexity's multi-model routing system?
Gemini serves as the primary heavy-lifting component that handles the core research tasks within Perplexity's system. While other specialized models contribute their individual strengths to specific subtasks, Gemini performs the main investigative work that drives the overall search and research process.
What are the key technologies that power Perplexity's deep-research system?
Perplexity's system runs on two main technologies: an Agent Search SDK and a framework called Search as Code. These technologies enable the system to break down queries into manageable pieces, route them to appropriate models, and orchestrate the entire research process automatically.
How does Perplexity ensure transparency and auditability in its search results?
Perplexity's system makes investigation automatic and auditable by attaching footnotes to each claim and logging every source used in the research process. This approach transforms search from simply answering a question into building a fully documented case with complete source attribution and evidence trails.
Why does Perplexity use multiple AI models instead of relying on a single advanced model?
According to Perplexity's philosophy, one AI model isn't smart enough to handle complex research tasks comprehensively, so they use twenty specialized models instead. This model orchestra approach allows each AI to excel at its specific domain while the system acts as a project manager coordinating all the different pieces of research.
Further Reading
- Introducing Perplexity Deep Research — Perplexity Blog
- Gemini Deep Research Agent | Gemini API — Google AI for Developers
- Gemini Deep Research — your personal research assistant — Google Gemini
- How OpenAI, Gemini, and Claude Use Agents to Power Deep Research — ByteByteGo Blog
- The Rise of Agent-Based Deep Research: Exploring OpenAI's Deep Research, Gemini Deep Research, Perplexity Deep Research, Ai2 ScholarQA, STORM, and More — Aaron Tay Substack
