Most Perplexity Users Don't Know It Can Run 19 AI Models at Once
Everyone treats Perplexity like a single smart search box. Hidden inside is Computer—a multi-agent system that dispatches 19 AI models in parallel to take a project from question to finished deliverable.
01. What It Is
Perplexity Computer is a multi-agent orchestration system built into Perplexity that runs an entire project end-to-end instead of answering one question at a time. When you hand it a complex task, it doesn't rely on a single model—it coordinates up to 19 different AI models working in parallel, each assigned to what it does best.
The system uses Claude Opus 4.6 as its core reasoning engine, then routes subtasks to specialists: Gemini for deep research, GPT-5.2 for long-context work, Grok for speed, Veo 3.1 for video generation, and Nano Banana for images. Rather than you switching tools and stitching results together by hand, Computer spins up sub-agents dynamically and runs them inside a cloud-based isolated sandbox that has a real filesystem, a browser, and more than 400 app integrations like Gmail, Slack, and Notion.
Because it runs asynchronously, a task can execute for hours—or keep running in the background over much longer stretches—without you babysitting it. It launched February 25–26, 2026, and remains active as of July 2026. The surprise is that despite being available, most people still fire off single-model queries and never touch the orchestration layer underneath.
Why It Matters
Instead of manually bouncing between a research tool, a document editor, an image generator, and a video tool, you describe the outcome once and let 19 models split the work in parallel. That collapses a multi-hour, multi-app workflow into a single conversation—and because tasks run asynchronously in the cloud, you can close your laptop and come back to a finished deliverable.
Who Can Benefit
- Analysts and researchers who need multi-source deep dives compiled into a report
- Founders and consultants juggling research, writing, and asset creation in one shot
- Marketers producing copy, images, and video for a campaign together
- Power users tired of copy-pasting between separate AI tools
02. Step-by-Step Guide
- 1
Open the Computer entry point
From the Perplexity Home page, click the Computer icon and choose Start a task. It's available on web, iOS, Android, macOS, and through Microsoft 365 app integrations.
- 2
Describe the full outcome, not just a question
Instead of asking one question, state the whole project—e.g. 'Research the top 5 competitors, summarize their pricing, and draft a comparison report with charts.' Computer breaks this into subtasks automatically.
- 3
Let it route models automatically
By default the system assigns each subtask to the best-fit model: Gemini for research, GPT-5.2 for long context, Veo 3.1 for video, Nano Banana for images, with Claude Opus 4.6 coordinating.
- 4
Override routing when you want control
If you'd rather force a specific model for a specific subtask, manually select it. This is useful when you trust one model's style or speed over the automatic pick.
- 5
Connect the apps it should touch
Authorize integrations (Gmail, Slack, Notion, and 400+ others) so Computer can pull inputs and deliver outputs directly into your tools inside its sandbox.
- 6
Walk away—it runs asynchronously
Tasks execute in the background for hours without intervention. Check back later for the completed deliverable rather than watching it work.
Pro Tips
- Frame requests as deliverables ('produce a report with charts and a summary email') so Computer knows the full pipeline, not just the first step.
- Use manual model override for subtasks where you have a strong preference—e.g. force Grok for speed-sensitive steps or Veo 3.1 for the video segment.
- Set up your key app integrations before your first big task so outputs land straight in Notion, Gmail, or Slack.
- Because tasks run in an isolated sandbox with a real browser and filesystem, treat it like a virtual coworker—give it clear success criteria.
Warnings & Limitations
- Computer is available exclusively to Perplexity Max subscribers ($200/month); the Pro tier only has limited access.
- It runs on a credit-based system—if credits run out, active tasks pause rather than cancel, but they won't finish until you top up.
- Model lineup and counts can shift; some sources note up to 20 models, and Perplexity may adjust which models handle which tasks.
- Long-running background tasks still need well-defined instructions—vague prompts can produce a lot of work in the wrong direction.
Related Tricks
Most ChatGPT Users Don't Know It Can Work on a Schedule—While You Sleep
ChatGPT isn't just for live chats. Its Tasks feature quietly runs recurring prompts on a schedule and pings you when the results are ready—no code required.
Most People Don't Know Gemini Can Keep Working After You Lock Your Phone
Gemini Spark is a cloud-based AI agent that runs your custom workflows 24/7 — triaging email, scheduling, and tracking topics even while you sleep. Here's how it turns Gemini from a chatbot into a background worker.
Most Perplexity Users Don't Know It Can Run 3 AI Models At Once—And Show Where They Disagree
Instead of betting on a single AI and hoping it's right, Perplexity's Model Council fires your question at three frontier models in parallel, then a fourth model synthesizes the answers and flags exactly where they agree and clash.