Perplexity Launches Computer, Expanding Access To Multiple AI Models
Perplexity has introduced a new product called Computer, an AI agent designed to carry out tasks by coordinating multiple AI models under one subscription.
The new offering positions Perplexity beyond its core AI search product and into software that can execute work on a user’s behalf. In reports about the launch, Computer is described as an agent that can run other agents, effectively acting as a manager that routes parts of a job to different systems.
Perplexity’s Computer coordinates 19 models, according to coverage of the product, and is priced at $200 per month. The product is being framed as a “computer” that can take instructions and then orchestrate the steps needed to complete them using a range of underlying AI capabilities.
The shift matters because it reflects an increasingly common approach in AI products: using multiple models rather than relying on a single system for every task. By packaging orchestration into a paid product, Perplexity is making a direct bet that users will value a single interface that can choose among different models and tools as work moves from asking questions to completing multi-step tasks.
Computer also underscores the industry’s push toward agentic AI, where systems do more than generate text responses. Products described as agents are meant to plan, delegate, and execute, which can change how people interact with software. Instead of toggling between separate chatbots, apps, and specialized tools, users are being offered a layer that can coordinate those pieces in the background.
The launch places Perplexity into a more crowded and competitive part of the market, where companies are racing to define what an AI “worker” looks like and how it should be priced. A $200-a-month subscription is a premium tier by consumer standards, suggesting Perplexity is targeting power users and professional workflows that can justify the cost.
Safety and control are also part of the conversation around agent products. Coverage has raised comparisons to other agent projects and discussed how Perplexity’s approach works. As agents get more capable, questions about permissioning, reliability, and guardrails become central to whether users and organizations will adopt them for real work.
What happens next will depend on how Computer performs in practice and how Perplexity evolves the product. The company will need to show that coordinating 19 models produces consistently better results than a single-model experience, and that users can trust an agent to carry out tasks without creating new complexity.
The debut of Computer signals Perplexity’s continued move from answering questions to executing work, and it raises the stakes for what users will expect from AI tools in the months ahead.
