What it does
Perplexity Computer is presented as a general-purpose digital worker that operates the same software interfaces a person would. You describe an outcome, and it breaks that into tasks and subtasks, creating sub-agents for work such as web research, document generation, data processing or API calls to your connected services. The work runs asynchronously, so you can do something else or run dozens of Computers in parallel. When it hits a problem it creates more sub-agents to solve it, and it checks in with you only if it truly needs you. Each task runs in an isolated compute environment with a real filesystem, a real browser and real tool integrations, with no local setup. The system is model-agnostic. The launch post says one model handles core reasoning while others are used for deep research, images, video, lightweight tasks and long-context recall, and that you can choose specific models for specific subtasks. The announcement is dated 25 February 2026 and says Computer is available to Perplexity Max subscribers, with Enterprise Max users to follow soon.
Capabilities
- Describe an outcome and it splits the work into tasks and sub-agents
- Runs workflows asynchronously, for hours or even months
- Orchestrates several AI models and lets you choose models for subtasks
- Isolated environment with a real filesystem, browser and tool integrations
- Run dozens of Computers in parallel
When to use it
- Multi-step projects that mix research, documents, data processing and calls to connected services
- Long-running work you want to hand off and check back on later
- Perplexity Max subscribers who want one system that draws on several top models
