Updated
Updated · ZDNet · Aug 25
Perplexity Unveils Portable Computer Local AI Agent, Requiring 24GB RTX GPUs for Windows
Updated
Updated · ZDNet · Aug 25

Perplexity Unveils Portable Computer Local AI Agent, Requiring 24GB RTX GPUs for Windows

2 articles · Updated · ZDNet · Aug 25

Summary

  • Portable Computer debuted Tuesday as a local-first version of Perplexity’s Personal Computer agent, keeping models, files and most task execution on the user’s device instead of in the cloud.
  • Local processing is meant to cut latency, reduce security exposure and avoid cloud inference charges, with credits used only when a task needs to go online or connect to services like Gmail, Slack or GitHub.
  • Paid access is limited to Pro, Max and enterprise tiers, and support is narrow for now: Linux first, with Windows promised in September.
  • Hardware is the main constraint—Linux users need an Nvidia RTX GPU or DGX Spark setup, while Windows PCs must have an Nvidia RTX GPU with at least 24GB of VRAM.
  • Portable Computer launches with Qwen 3.8 27B and Perplexity’s post-trained PPLX 27B, with Nvidia’s Nemotron 3.5 Lightning slated to follow as Perplexity pushes local AI beyond its February cloud-based release.

Insights

Could Perplexity’s shift to high-end local hardware signal the end of cloud-dependent AI for enterprise users?
Does moving complex AI tasks to personal devices create hidden security risks that cloud platforms already solved?
Will the massive 24GB VRAM requirement make true AI privacy a luxury only the wealthy can afford?