perplexity, Portable ComputerIt brought native artificial intelligence tasks to computers by adding NVIDIA RTX support to the Windows application. For local use, your GeForce RTX or RTX PRO card at least 24 GB memory needs to carry it. So you can analyze your files on your computer and run multi-step tasks directly on your own hardware without sending sensitive content to the cloud. Purely local tasks do not consume Computer credits, but access requires a Pro or Max subscription.
When you compare the memory condition across cards, the GeForce RTX 5090 32GB GDDR7 memory exceeds the required minimum limit. In contrast, GeForce RTX 5080, 16GB GDDR7 With its memory, it does not meet the memory requirement in the RTX version of Portable Computer. The 24 GB threshold here refers to the video card’s own memory, that is, VRAM capacity, not the system memory of the computer. The new generation RTX label alone is not enough; You can check the VRAM capacity of your card before installation.
Instead of just offering a search engine interface, Portable Computer examines files, breaks tasks down into steps, and completes them with tools. In these processes, the artificial intelligence model, task manager, scheduler and local search index run directly on your computer. Perplexity for these tasks Qwen 3.8 27B by putting the model through additional training PPLX 27B developed the model. Additionally, the application handles native model installation from within itself, so you don’t need to select or install separate software packages.
Perplexity first made Portable Computer available for Linux-based NVIDIA DGX Spark on August 25. The company later opened the system to GeForce RTX and RTX PRO computers running Linux. This Linux version also required at least 24 GB VRAM on the graphics card for local use. With Windows support, you no longer need to switch to Linux to run native tasks.
Portable Computer asks for approval before moving to the cloud
The Windows application supports Outlook, OneDrive and Word connections, as well as Google Drive, Gmail, Slack and GitHub connections. For example, you might want GitHub to review open pull requests in your project, evaluate code changes, and extract next steps for each request. In addition, through local data, the application can analyze at what stage users leave after registration before reaching their first task. It then pushes the results of this analysis directly to the team’s Slack channel. Use the built-in browser when executing tasks, and use the built-in browser for code operations. SPACE It uses the protected workspace named .
When up-to-date information or further reasoning is needed, the Portable Computer can take help from Perplexity search and cloud models. However, it asks for approval before sending content from the device and only shares information you approve. The advisor model in the cloud does not gain direct access to your files and local tools when working with this context. Instead, it directs the local model via text, which carries out the task, and leaves the processing to tools on the computer. You can turn off web search and run tasks completely offline when you’re only working with your own files.
To begin installation, you can download Perplexity’s Windows app from the Microsoft Store, then sign in with your Pro or Max account. Local use requires compatible hardware as well as a subscription. Tasks you complete entirely on the computer do not reduce your computer credits. The steps where you get help from cloud models bring additional costs beyond this local use.
We did not test the Windows version for this news; The performance figures come from Perplexity’s work on DGX Spark. The company’s Local Knowledge Work Bench rating represents day-to-day knowledge work. 53 missions Contains. In this test PPLX 27B with Computer 85.4 percentComputer, which runs on the same hardware as the base Qwen 3.8 27B model, scored 82.6 percent. Pi, alternative task execution software that works with the same Qwen model, reached 77.6 percent and Hermes reached 74.0 percent. The company ran each task three times in this evaluation. Therefore, the measurements do not give a direct acceleration rate for RTX graphics cards on Windows.
perplexity, Today HP made a separate Windows announcement and announced support for the HP ZBook Ultra G3a. HP will start the deployment with this computer and pre-install the Perplexity Windows application. this computer AMD Ryzen AI Max PRO 400 series processors, but Portable Computer access varies depending on the hardware option on each device. NVIDIA’s 24 GB VRAM requirement for RTX cards does not constitute a common Windows requirement that also covers HP’s computers with AMD processors.