Perplexity has launched Computer, a product that turns a single user prompt into a coordinated swarm of AI agents working across 19 different models. It's the most aggressive bet yet that the future of AI isn't about picking the best model — it's about orchestrating all of them.
How It Works
Users enter one prompt. Computer automatically decomposes it into smaller tasks and subtasks, then spawns sub-agents to handle each piece: research, document creation, web browsing, data processing, and more. Each sub-agent can select the most appropriate model for its specific task from Perplexity's pool of 19 integrated models.
The result is a system that handles complex workflows — market research reports, competitive analyses, multi-source data synthesis — that would normally take hours of manual work across multiple tools. It's available on Perplexity Max ($200/month), positioning it as a premium productivity tool for professionals.
The Multi-Model Thesis
This launch doubles down on the thesis behind Perplexity's recently introduced Model Council, which runs three frontier models simultaneously and compares outputs for higher-confidence answers. Computer extends that idea from answers to actions.
The approach directly challenges the single-model paradigm that companies like OpenAI and Anthropic have built their businesses around. Instead of trusting one model to do everything, Perplexity is betting that combining specialized models produces better results — and that users will pay a premium for the orchestration layer.
It also puts Perplexity in direct competition with autonomous agent platforms like OpenClaw and the growing ecosystem of AI coding agents that handle multi-step workflows.
What Else Shipped
Perplexity packed several other updates into the same release cycle:
- Deep Research upgrade — now runs on Claude Opus 4.5 for Max and Pro users, achieving state-of-the-art performance on external benchmarks
- Learning Mode — expanded beyond students to all users, offering step-by-step explanations
- In-chat shopping — a PayPal and Venmo integration that lets users go from product research to payment without leaving the conversation
The shopping integration is notable because Perplexity recently abandoned advertising entirely. Commerce — taking a cut of transactions — may be the alternative revenue stream that makes the no-ads model sustainable.
The $200/Month Question
Computer's pricing will filter out casual users, but that's the point. At $200/month, Perplexity is competing with enterprise tools, not consumer chatbots. The question is whether multi-model orchestration delivers enough value over a single frontier model to justify the premium.
For developers interested in building similar multi-agent systems, understanding the MCP protocol is becoming essential — it's the connective layer that makes tools like Computer possible. FreeAcademy's MCP course covers building these integrations from scratch.
Early reports suggest Computer handles research and analysis tasks well but struggles with highly creative or ambiguous requests where model disagreement produces incoherent results. It's a v1 — impressive in scope, uneven in execution. But the direction is clear: AI is moving from single-model chat to multi-model orchestration, and Perplexity is leading the charge.
