Perplexity is bringing more AI work directly onto your PC. Its Portable Computer agent is now available for Windows users with compatible Nvidia RTX hardware, giving people a way to run multi-step AI tasks locally instead of sending everything to the cloud.
There is one major catch: the system requires an Nvidia GeForce RTX or RTX PRO GPU with at least 24GB of VRAM. That puts the feature firmly in high-end desktop and workstation territory for now.
What is Perplexity Portable Computer?
Portable Computer is the local version of Perplexity’s Computer agent. Instead of acting like a basic chatbot that only answers one prompt at a time, it is designed to work through longer tasks, use tools and complete multiple steps on your behalf.
The important difference is where the AI runs. With Portable Computer, much of the workload can stay on your own machine. That can reduce dependence on cloud processing and may be useful for people who want faster local workflows or more control over where their data is processed.
Windows support is the big update
Portable Computer first appeared for Linux systems. Perplexity and Nvidia have now expanded the experience to Windows through the Perplexity app, opening it to a much larger group of PC users.
Nvidia had already previewed broader local-agent support at IFA 2026, saying that Perplexity Portable Computer, Hermes Agent and OpenClaw were part of a push to make local AI easier to install and run on RTX systems.
The 24GB VRAM requirement will exclude most PCs
The biggest limitation is graphics memory. A gaming PC can be very powerful and still fall short if its GPU has less than 24GB of VRAM.
Examples of GeForce cards that meet the stated memory threshold include the RTX 3090, RTX 3090 Ti and RTX 4090 with 24GB, as well as the RTX 5090 with 32GB. Nvidia’s professional RTX PRO cards with 24GB or more also fit the requirement. Meeting the VRAM threshold does not necessarily mean every configuration has been individually validated, so users should still check Perplexity’s current compatibility information before upgrading hardware.
Why run an AI agent locally?
- Privacy: more work can stay on your own PC instead of being processed remotely.
- Local performance: a powerful GPU can handle AI inference without waiting on a cloud service for every step.
- Fewer cloud limits: local tasks do not depend on the same type of usage credits as fully cloud-based workloads.
- Desktop integration: Perplexity says the Windows version supports local MCP servers and scheduled recurring tasks, which can make the agent more useful for real workflows.
The agent can still use cloud models when needed and when the user allows it, so this is not an entirely offline replacement for every AI service. Think of it more as a hybrid setup that can keep suitable tasks local while still reaching the cloud for heavier work.
What kind of PC do you need?
For most users, the GPU is the deciding factor. The requirement of 24GB or more of Nvidia VRAM means many popular RTX gaming cards are not eligible, even if they are relatively new.
Perplexity also requires a paid Pro or Max subscription for Portable Computer. If you already own a high-memory RTX desktop, however, the new Windows support gives you a way to experiment with local agent-style AI without buying a separate dedicated AI machine.
This could be the start of a bigger Windows local-AI push
Nvidia is clearly trying to make local AI agents easier for regular PC users. The company has also announced RTX Spark Windows systems and simpler local setups for other agent platforms. If these tools become easier to install and the hardware requirements come down over time, local AI could become a normal part of high-end Windows PCs rather than something limited to developers and enthusiasts.
For now, though, Portable Computer is best viewed as an early look at that future. The software is becoming easier to access, but the 24GB VRAM requirement means the hardware barrier is still high.
Sources
Tom’s Hardware reported the Windows release and current hardware requirement on September 15, 2026. Nvidia previously detailed its local-AI plans and Portable Computer support in its IFA 2026 local AI announcement.