Perplexity adds Portable Computer to Windows RTX PCs

Perplexity has added Portable Computer to its Windows app for compatible NVIDIA RTX PCs, extending its local AI software to more Windows users.

The software lets users run an AI agent on-device for multistep work such as analysing files, pulling together information across documents and handling routine tasks. Sensitive information can remain on the PC when work is completed locally, while more demanding tasks can be sent to cloud models with user permission.

Portable Computer is now available on Windows machines with compatible NVIDIA GeForce RTX GPUs and NVIDIA RTX PRO workstations with 24GB or more of VRAM. The launch builds on existing support for NVIDIA DGX Spark systems and RTX PCs running Linux.

The rollout reflects a broader industry push to move more AI processing from remote data centres to local hardware, particularly for customers concerned about privacy, latency and the cost of repeated cloud use. Perplexity's approach combines local models with cloud-based systems in one application.

Under the company's setup, the Windows app includes a local model tuned for Perplexity Computer and optimised for NVIDIA RTX GPUs. That removes the need for users to choose models themselves or configure the software tools often required to run AI models on a personal machine.

The software can use built-in tools including a browser and what Perplexity calls its SPACE sandbox. It also connects with workplace applications and file services including Microsoft Outlook, OneDrive, Word, Google Drive, Gmail, Slack and GitHub.

Those integrations are intended to support common office and technical workflows. In software development, for example, the agent can review open pull requests in a GitHub project, organise them by status, identify next steps and flag outdated documentation before submitting proposed updates for review.

Perplexity also outlined financial and startup use cases. In one example, the software can examine brokerage summaries, tax returns and other records to identify recurring holdings linked to fees and tax drag while keeping documents on the device. In another, it can analyse a startup's funnel export locally to identify where new signups drop off between installation and the first completed task, then post findings to Slack.

Local-first push

The emphasis on local execution comes as AI companies try to persuade businesses and individual users that agents can take on more complex work without forcing all data into external systems. By keeping processing on the machine where possible, vendors aim to address confidentiality concerns while reducing reliance on paid cloud inference.

Work completed locally does not consume Perplexity Computer credits. That suggests the company is drawing a commercial distinction between tasks run on the user's own hardware and those escalated to remote models for higher-level reasoning or research.

NVIDIA has been promoting the idea that newer PC and workstation GPUs can support increasingly capable local AI workloads, and Perplexity's Windows launch adds another software partner to that effort. The 24GB-or-more VRAM requirement, however, means the software targets higher-end systems rather than the broader consumer PC base.

Hardware limits

That hardware threshold may shape the product's near-term market. High-memory RTX cards are common in professional desktops and some enthusiast PCs, but they remain out of reach for many mainstream laptop and office users, limiting immediate adoption even as interest in local AI tools grows.

Even so, the launch highlights how AI agents are moving beyond chatbot-style question answering towards software that can inspect documents, navigate connected applications and complete sequences of work on behalf of users. The combination of local processing for sensitive material and optional cloud handoff for harder tasks is becoming a common design choice as suppliers try to balance privacy with model performance.

Portable Computer is available on Windows systems with supported NVIDIA GeForce RTX and RTX PRO GPUs with 24GB or more of VRAM.