Superpower Daily

Superpower Daily

Your reading space

Sign in
Bolt.new Adds Forge, Offering 50x More AI Coding Usage for Opt-...Manhattan DA Seizes 12 Domains Accused of Selling Celebrity Dee...OpenAI Uses Contractors to Review Real ChatGPT Prompts, 404 Med...Anthropic Tells Investors It Expects Another Quarter of Positiv...UK Rights Committee Calls for AI Law and Regulator With Enforce...MIT Researchers Publish HardFlow to Keep AI Outputs Within Hard...Bolt.new Adds Forge, Offering 50x More AI Coding Usage for Opt-...Manhattan DA Seizes 12 Domains Accused of Selling Celebrity Dee...OpenAI Uses Contractors to Review Real ChatGPT Prompts, 404 Med...Anthropic Tells Investors It Expects Another Quarter of Positiv...UK Rights Committee Calls for AI Law and Regulator With Enforce...MIT Researchers Publish HardFlow to Keep AI Outputs Within Hard...
Swipe

Perplexity Adds a Local AI Agent to Windows PCs With 24GB RTX GPUs

The new Windows version keeps file analysis on-device and can seek approval for a cloud handoff, but it is limited to high-memory Nvidia systems.

By 2 min read
Perplexity Adds a Local AI Agent to Windows PCs With 24GB RTX GPUs
Perplexity Adds a Local AI Agent to Windows PCs With 24GB RTX GPUs

Listen to this story

The audio brief

About 1:25
0:001:25
Read transcript
Perplexity’s Portable Computer agent is now available on Windows, but only on machines with an Nvidia GeForce RTX or RTX PRO graphics card carrying at least 24 gigabytes of video memory. That gives users a way to run multistep AI work on local files and connected services, instead of sending every task to the cloud from the start. The agent uses a local model, such as Qwen 3.8 27B, to analyze information on the PC. It can work across Outlook, OneDrive, Word, Google Drive, Gmail, Slack, and GitHub, handling tasks involving documents, communications, and code. Nvidia gives examples including reviewing GitHub pull requests, analyzing brokerage and tax documents, or finding signup-funnel trends and sending the results to Slack. The important distinction is that this is local-first, not fully offline. If the agent decides a task needs stronger reasoning, it can ask the user for permission to send information to cloud models. Nvidia also says locally completed work does not use Perplexity Computer credits. The hardware requirement is the catch. Windows systems with less than 24 gigabytes of VRAM are excluded, so this launch targets high-memory Nvidia PCs. The approach already works on Linux RTX machines and DGX Spark systems, while DGX Station support is planned. For now, the key question is simply whether a Windows PC has enough GPU memory to qualify.

Story brief

3 key points

Perplexity’s Portable Computer is now available on Windows PCs equipped with Nvidia RTX or RTX PRO GPUs carrying at least 24GB of VRAM. The agent runs Qwen 3.8 27B locally to work across files and connected services, including GitHub, Outlook, Slack, and Google Drive, without using Perplexity Computer credits. It is local-first rather than fully offline: users must approve cloud escalation when a task needs stronger...

  1. 01

    Supported hardware includes GeForce RTX and RTX PRO cards; lower-memory Windows GPUs are excluded.

  2. 02

    Connectors cover Outlook, OneDrive, Word, Google Drive, Gmail, Slack, and GitHub.

  3. 03

    Portable Computer can request permission to send information to cloud models when local reasoning is insufficient.

Perplexity has brought its Portable Computer agent to compatible Windows PCs, letting users run multistep AI work on local data rather than beginning in the cloud. The release offers a privacy-minded workflow for files and connected services, but only on Nvidia RTX systems with at least 24GB of video memory.

Portable Computer is the local version of Perplexity Computer, an agent designed to plan and carry out multistep tasks on a PC. In the Windows app, it can use local models to analyze data and files while keeping sensitive information on the device.

A private workflow, with an approved handoff

The product is not an always-offline system. When Portable Computer identifies a task that needs more advanced reasoning, it can ask for permission before sending information off-device to cloud models. Nvidia also says work completed locally does not consume Perplexity Computer credits.

A high-memory entry requirement

The Windows release supports Nvidia GeForce RTX and RTX PRO GPUs with 24GB or more of VRAM, the dedicated memory used by a graphics processor. Nvidia says the app simplifies local setup with a model such as Qwen 3.8 27B, post-trained for Perplexity Computer and optimized for RTX GPUs.

Connected to the tools where work already lives

Portable Computer includes connectors for Microsoft Outlook, OneDrive, Word, Google Drive, Gmail, Slack and GitHub. That gives the agent access to documents, communications and code repositories within the same local-first setup.

Nvidia’s example uses

  • Review connected GitHub pull requests, organize them by status and identify next steps.
  • Analyze brokerage summaries and tax documents for holdings creating fees and tax drag.
  • Examine signup-funnel data and send the top findings to Slack.

The Windows availability extends an approach already supported on Nvidia DGX Spark systems and RTX PCs running Linux. Nvidia says support for DGX Station is expected later; for now, the key limitation is whether a Windows machine clears the 24GB memory threshold.

Sources

  1. blogs.nvidia.comPerplexity Portable Computer Is Now Available on Windows, Powered by NVIDIA RTX

Loading discussion...

Related coverage

Continue reading