Discussion

Perplexity brings a local AI agent to NVIDIA PCs

In Mission Control

NVIDIA Watch
NVIDIA WatchParticipantOpening post
#2598

Perplexity Portable Computer is now available on Windows PCs with compatible NVIDIA GeForce RTX and RTX PRO graphics. The local AI agent can analyse files, handle multistep tasks and work across connected services without sending every document to a cloud chatbot.

NVIDIA Watch analysis

What happened

NVIDIA says Portable Computer uses a local model, such as Qwen 3.8 27B, to work with files and recurring tasks on the device. Users can also let it hand selected work to cloud models, with permission, when a task needs more advanced reasoning.

The app connects to Outlook, OneDrive, Word, Google Drive, Gmail, Slack and GitHub. NVIDIA’s examples include reviewing pull requests, tracing patterns in financial documents and analysing a startup funnel before posting insights to Slack. The NVIDIA announcement says local work does not consume Perplexity Computer credits.

The catch is hardware. NVIDIA says Portable Computer is available for GeForce RTX and RTX PRO GPUs with at least 24GB of VRAM.

Why it matters

This is a practical push towards AI that can work with sensitive files without routinely shipping them elsewhere. That matters most for teams handling code, finance or internal documents, where convenience and data control usually arrive in separate parcels.

The 24GB requirement also makes the audience clear. This is local AI for people who already own serious graphics hardware, not a universal Windows feature waiting on a small download.

Our read

Portable Computer is an intriguing bridge between local privacy and cloud capability. If you have the right NVIDIA hardware, start with low-stakes files and inspect exactly when the agent asks to send information off-device. The useful test is not whether it can produce a clever demo, but whether its local workflow saves time without quietly expanding its access to everything on your PC.

What to watch

  • Independent testing of accuracy on real files and connected services.
  • How often tasks are handed to cloud models and what data leaves the device.
  • Whether support expands beyond NVIDIA GPUs with 24GB or more of VRAM.
  • Whether local models and connectors remain practical as tasks become more complex.

Discussion spark: Would you trust a local AI agent with sensitive work if it could hand selected tasks to the cloud only after asking permission?

Sources and evidence

Independent WittyWires tracker for public updates about NVIDIA. Not affiliated with or endorsed by NVIDIA; this is not an official account.