Perplexity Launches Portable Computer for NVIDIA RTX PCs

Perplexity is bringing Portable Computer to Windows, giving users a way to run its agentic AI system locally on compatible NVIDIA RTX PCs.

Portable Computer is a local version of Perplexity Computer that can plan and execute multistep tasks directly on a PC. Running models locally also lets users work with sensitive data without automatically sending it to the cloud, while reducing the use of cloud credits.

Portable Computer combines local and cloud AI

Portable Computer can analyze files, combine information from different sources, and run recurring workflows using local AI models.

For tasks that require more advanced reasoning, the agent can turn to cloud-based models. Perplexity says it asks for permission before sending information off the device.

Perplexity Computer also includes a built-in browser and the company’s proprietary SPACE sandbox for handling tasks in a controlled environment.

NVIDIA RTX GPUs power local AI workloads

Portable Computer can run models such as Qwen 3.8 27B, which Perplexity has post-trained for its Computer platform and optimized for NVIDIA RTX GPUs.

The setup removes much of the manual work normally required to choose models and configure a local AI software stack.

Portable Computer requires an NVIDIA GeForce RTX or RTX PRO GPU with at least 24GB of VRAM. Support for NVIDIA DGX Station is also planned.

Connectors bring Outlook, Gmail, Slack, and GitHub into the agent

Portable Computer supports integrations with Microsoft Outlook, OneDrive, Word, Google Drive, Gmail, Slack, and GitHub.

These connectors allow the agent to combine local processing with files, messages, repositories, and services that users already rely on.

For example, Portable Computer can review GitHub pull requests, analyze sensitive financial documents locally with source citations, or examine business datasets before sharing selected insights through services such as Slack.

Perplexity pushes more agentic AI onto the PC

The Windows launch gives Perplexity another way to run increasingly capable AI agents without requiring every task or dataset to leave the user’s computer.

The local-first approach also gives users more control over when information gets sent to cloud-based models.

In other news, NVIDIA confirmed an October launch for RTX Spark, while the company could also launch a new RTX 5070 with a GB203 GPU.