What is Perplexity Computer and how does the AI digital worker use multiple AI models to get work done?

Perplexity Computer explained: For years, artificial intelligence has mostly lived inside chat boxes, answering questions, drafting emails, summarizing documents. But as AI models grow more powerful, that format is starting to feel limiting.

What Is Perplexity Computer

On Wednesday, Perplexity unveiled Perplexity Computer, describing it not as another chatbot or standalone agent, but as a general-purpose digital worker designed to operate the same interfaces humans use and to handle entire workflows from start to finish.

How Perplexity Computer Turns Outcomes Into Automated Workflows

Instead of asking a question and receiving an answer, users describe an outcome, as per Perplexity. The system then breaks that outcome into tasks and subtasks, automatically creating sub-agents to execute them.


One agent might gather research. Another drafts a document. Others process data or make API calls to connected services.

The work happens in the background, asynchronously. Users can move on to other responsibilities or even run multiple Perplexity Computers at once, while the system continues working.

Also read: Word of the day: Opal

How Perplexity Computer Uses Sub-Agents to Complete Tasks

If it encounters a challenge, it generates additional sub-agents to resolve it. It can research missing information, find API keys, build applications if needed, and only check in when human input is truly necessary. Each task runs in its own isolated compute environment with access to a real filesystem, a real browser, and integrated tools.

Perplexity describes that, "The result is a safe harness for powerful AI, universally accessible without localized setup, and delivering the best work of every current top AI model."

Also read: Why did NVDA stock drop after Nvidia earnings beat estimates? Here's why investors are worrying about long-term AI dominance outlook

Which AI Models Power Perplexity Computer

According to Perplexity, AI models are not becoming interchangeable, they are becoming specialized. Each frontier system excels at different kinds of work, meaning a full workflow requires access to multiple models deployed intelligently.

Perplexity Computer runs Opus 4.6 as its core reasoning engine and coordinates sub-agents with different models depending on the job:

• Gemini for deep research
• Nano Banana for image generation
• Veo 3.1 for video
• Grok for lightweight, speed-focused tasks
• ChatGPT 5.2 for long-context recall and broad search

The system is model-agnostic, meaning integrations can change as models advance. Users can also select specific models for individual subtasks, offering choice and control as token budgets become a more common consideration.

FAQs

Does Perplexity Computer work on its own?
Yes. Once you describe the outcome, it works asynchronously and only checks in if truly needed.

Can Perplexity Computer run multiple tasks at once?
Yes. Users can even run multiple Perplexity Computers in parallel.