Perplexity Computer is introduced as a novel AI system designed to unify disparate top-tier AI capabilities into a single, general-purpose digital worker. Positioned as the next evolution of AI beyond chat interfaces or simple agents, it is capable of creating and executing complex workflows that can span hours or even months.

The core methodology of Perplexity Computer involves operating the software stack in a manner akin to a human co-worker. Users initiate a process by describing a desired outcome. Perplexity Computer then internally decomposes this outcome into a hierarchy of tasks and subtasks. For the execution of these granular tasks, it dynamically creates and manages specialized sub-agents. These sub-agents are capable of performing a diverse range of operations, including but not limited to web research, document generation, data processing, and making API calls to user-connected services. The system features automatic coordination among these sub-agents, enabling asynchronous work and the potential for parallel execution of multiple Perplexity Computer instances. A critical aspect of its operational robustness is its ability to self-correct: when encountering a problem, it automatically spawns new sub-agents dedicated to solving that specific issue, which may involve tasks like retrieving API keys, conducting supplementary research, or even generating code for applications. Each task is executed within an isolated compute environment, providing secure access to a real filesystem, a real web browser, and various real-world tool integrations, ensuring both security and practical utility without requiring localized setup.

Perplexity Computer is presented as a natural evolution of Perplexity's mission to power curiosity through highly accurate AI. Building upon foundations established with products like Comet (an AI-native browser) and Comet Assistant (a personal AI agent), Perplexity has historically focused on deep research for accuracy, incorporating persistent memory, and task management. A fundamental principle underpinning Perplexity Computer is its model-agnostic architecture. Recognizing that frontier AI models are increasingly specializing rather than commoditizing, the system is designed to orchestrate and leverage the best models for specific tasks. For instance, at launch, Perplexity Computer utilizes Opus 4.6 as its core reasoning engine, while intelligently delegating tasks to other specialized models: Gemini for deep research, Nano Banana for image generation, Veo 3.1 for video processing, Grok for speed in lightweight tasks, and ChatGPT 5.2 for long-context recall and extensive search. This modular design allows for the seamless integration of new and improved models as they emerge, offering users control over model selection for specific subtasks.

The nomenclature "Computer" is drawn from its historical meaning, referring to human apprentices who performed complex, specialized calculations to achieve accurate results through autonomous division of labor, as exemplified by Alexis Clairaut's work on Halley's Comet. Perplexity Computer aims to bring this concept to AI, orchestrating diverse capabilities across tasks, tools, and time. It is currently available to Perplexity Max subscribers and will soon be extended to Enterprise Max users.