Perplexity has announced “Perplexity Computer,” a new agentic platform that combines files, tools, memory, and a suite of 19 AI models into a single orchestrated system aimed at executing entire projects from research to deployment.
Positioned as a “general‑purpose digital worker,” Computer is built to run long, autonomous workflows in the cloud so users can delegate multi‑day or even month‑long tasks.
A Multi‑Model AI Orchestra
At the heart of Perplexity Computer is a multi‑model architecture where different AI systems are treated as specialized components rather than standalone chatbots.
Claude Opus 4.6 serves as the core reasoning engine, while other frontier models such as Gemini, Grok, and ChatGPT 5.x are dispatched for focused tasks like coding, writing, or data extraction.
CEO Aravind Srinivas likens this to conducting an orchestra: “Musicians play their instruments, I play the orchestra,” emphasizing that no single model family can deliver optimal results without the talents of others.
This orchestration layer routes subtasks to the most capable model for that specific job, enabling what the company calls “sophisticated token management.”
Users can even lock certain subtasks to specific models, giving power users granular control over quality, speed, and cost.
From Search to Execution: What Computer Can Do
Perplexity Computer is designed as more than an upgraded chatbot; it aspires to function as a full “cloud computer” that can research, analyze, design, code, deploy, and then continuously manage projects. The system integrates:
- A file system for storing and organizing project assets.
- CLI-style tools and a sandboxed execution environment for running code and automations safely.
- A browser with real‑time internet access for live research and data gathering.
- Connectors to personal and enterprise data sources, allowing workflows that tie into existing stacks and CRMs.
Internally, Perplexity has already used Computer to build large artefacts such as a 4,000‑row spreadsheet generated overnight, showcasing how the platform can be kept running for extended periods on complex, data-heavy workflows.
Pricing, Credits, and Availability
Perplexity Computer is launching first for Perplexity Max customers, the company’s top‑tier subscription priced at 200 dollars per month.
Max subscribers receive 10,000 credits each month along with a one‑time 20,000‑credit launch bonus, valid for 30 days, to encourage heavy real‑world testing of the new system.
This marks Perplexity’s first move into usage‑based, per‑token billing on the consumer side, which Srinivas frames as a more sustainable business model for AI than ad‑supported experiences.
The company plans to open Computer to Pro and Enterprise tiers once it is satisfied with load testing and real‑world reliability.
For now, access is available via the web interface and requires users to be signed in with an eligible account.
A Busy Product Cycle for Perplexity
The debut of Perplexity Computer comes during an intense period of shipping for the company, as it pivots from being an AI search challenger to a broader execution platform.
Recently, Perplexity has upgraded its Deep Research feature to run on Anthropic’s Claude Opus 4.6, rolled out a “Model Council” capability that queries multiple frontier models at once, and added more models from partners like Anthropic and Google.
Perplexity’s AI assistant is also expanding its footprint in consumer hardware: the company has confirmed that its Comet AI browser is headed to the iPhone, while Samsung is preparing to integrate Perplexity’s assistant into the Galaxy S26 lineup, triggerable via the wake phrase “Hey Plex.”
With Computer, Perplexity is betting that the next wave of AI differentiation will come not from a single frontier model, but from how many models, tools, and data sources can be orchestrated into a cohesive, always‑on system that behaves more like a modern operating environment than a chatbot.
