The Pulse
Perplexity Brings Its Local Agent to Windows RTX PCs
Perplexity Portable Computer is now available on compatible Windows PCs with NVIDIA GeForce RTX and RTX PRO GPUs. The local agent runs private workflows on-device, while asking permission before sending information to cloud models.
AI.info Team ·
Portable Computer reaches Windows RTX systems
Perplexity’s local agent is now available on Windows PCs equipped with NVIDIA GeForce RTX or RTX PRO GPUs carrying at least 24GB of VRAM. The release brings Portable Computer, a local version of Perplexity Computer, to a much larger hardware base after its initial rollout on NVIDIA DGX Spark systems and RTX PCs running Linux.
Portable Computer is designed to plan and execute multistep tasks across local files, connected applications and the web. Work handled by the local model stays on the machine and does not consume Perplexity Computer credits, while tasks that need cloud-based research or more advanced reasoning can be sent off-device only after the user gives permission.
NVIDIA announced the Windows release on September 14, 2026, describing it as a way to run private AI workflows on compatible Windows hardware. Perplexity’s earlier product announcement described Portable Computer as a local-first version of its agent that runs its orchestrator, planner, tool router, scheduler and search index on the user’s machine.
Qwen 3.8 27B handles local work
The Windows application includes a local model such as Qwen 3.8 27B, which Perplexity post-trained to work with Computer and optimized for NVIDIA RTX GPUs. Users do not need to select a model, configure an inference server or assemble the supporting software stack themselves. The setup is intended to make local execution part of the same application rather than a separate developer project.
Portable Computer can analyze files, search across documents and code, and carry out actions on the device. NVIDIA says the application also includes Perplexity Computer’s built-in browser and proprietary SPACE sandbox, where code and tool execution run in isolated environments with controlled access to files and connected applications.
The local model cannot independently supply current information from the outside world. When a task needs web access, a connected cloud service or one of Perplexity’s more advanced models, the local orchestrator can request cloud support and ask the user to approve sending information away from the PC.
Microsoft and Google connectors widen the agent’s reach
Portable Computer connects to Microsoft Outlook, OneDrive and Word, as well as Google Drive, Gmail, Slack and GitHub. Those connections let the agent work across files, messages, code repositories and workplace communications without requiring users to copy material into a separate chatbot.
NVIDIA gives several examples of the intended workflow. An engineering team could have the agent review open pull requests, group them by status and identify next steps. A finance user could point it at brokerage summaries, tax documents and other records to trace recurring fees, with the results cited to the relevant file and page while the documents remain on the machine.
A startup could also ask the agent to examine a local funnel export, identify where new users stop progressing and post selected findings to Slack. Such actions still depend on the permissions granted to connected services, and cloud-bound transfers require approval when the task needs to move information beyond the computer.
Windows expands the audience beyond DGX Spark
Perplexity introduced Portable Computer first for NVIDIA DGX Spark, a compact system built around the Grace Blackwell GB10 platform with a 20-core Arm CPU and 128GB of unified memory. Its initial release ran on Linux and targeted Pro and Max subscribers. Windows support extends the product to owners of higher-memory GeForce RTX and RTX PRO hardware already used for gaming, creative work and local model inference.
The move also gives Perplexity a way to separate routine processing from paid cloud usage. Local analysis, file handling and other on-device tasks do not count against Computer credits, while cloud research remains available when local models lack web access or the reasoning capacity required for a particular job.
NVIDIA says support for DGX Station is expected next. For Windows users, the immediate requirement is a compatible RTX or RTX PRO GPU with at least 24GB of VRAM, with Portable Computer installed through the Perplexity application.