Perplexity has introduced Perplexity Computer, a system that combines multiple AI capabilities into a single platform. It is designed to research, design, code, deploy, and manage projects from start to finish.
Unlike standard chat interfaces that generate answers or task-based agents that complete isolated actions, Perplexity Computer creates and executes full workflows that can operate for extended durations.
What is Perplexity Computer?
Perplexity Computer operates digital tools in a manner similar to a human user. After a user defines an intended outcome, the system divides it into structured tasks and subtasks, assigns sub-agents, and executes them in parallel.
Sub-agents can:
- Perform web research
- Draft and edit documents
- Process and analyze data
- Make API calls to connected services
- Write, test, and deploy code
Workflows run asynchronously. Users can operate dozens of Perplexity Computer instances at the same time and scale from a single task to hundreds of active projects.
How It Works
Each task runs inside an isolated compute environment that includes:
- A real filesystem
- A real browser
- Integrated tools and external APIs
If the system encounters a problem, it generates additional sub-agents to address it. This may involve researching missing details, locating API keys, building applications when required, or requesting clarification only when necessary. The platform is accessible via the web and does not require localized installation.
Multi-Model Orchestration Across 19 Models
Perplexity Computer uses a model-agnostic orchestration framework that routes work to the model best suited for each subtask.
At launch, the system includes:
- Anthropic Claude Opus 4.6 as the core reasoning engine for task matching
- Google Gemini for deep research and sub-agent creation
- OpenAI ChatGPT 5.2 for long-context recall and broad search
- xAI Grok for lightweight and speed-focused tasks
- Google DeepMind Veo 3.1 for video generation
- Nano Banana for image generation
In total, the platform can route work across 19 different models. Users can manually select specific models for individual subtasks and apply spending controls to manage token usage. Because the architecture is model-agnostic, models can be updated as new versions become available.
Infrastructure and Memory
Perplexity Computer runs on Perplexity’s infrastructure and includes:
- Hundreds of service connectors
- Persistent memory that retains prior work
- File handling capabilities
- Web access
The system maintains context across projects while applying default security controls.
Pricing, Credits, and Availability
Perplexity Computer uses usage-based pricing and is available on the web for Max subscribers starting today.
Max subscribers receive:
- 10,000 credits per month included with the subscription
- A one-time bonus of 20,000 additional credits
- Bonus credits granted at launch for existing users and at signup for new users
- Bonus credits that expire 30 days after being issued
Usage controls include:
- Model selection for individual sub-agent tasks
- Spending caps to manage token usage
Support for Pro and Enterprise plans is expected to roll out soon.