NVIDIA and Perplexity Unveil Portable Computer Agent

NVIDIA and Perplexity have jointly launched Portable Computer, an Agentic AI platform designed to operate locally on NVIDIA DGX Spark hardware or via cloud models. The agent provides developers with a choice between local and cloud inference, enabling sensitive code, proprietary data, and project context to remain within their own system. The platform is optimized for NVIDIA RTX GPUs and includes support for the post-trained Qwen 3.8 27B model, with a fine-tuned Nemotron 3.5 Lightning variant planned for future release.

Key Facts

How Does the Portable Computer Agent Work?

Portable Computer operates as an agent harness combined with an orchestrator, offering users the option to switch between running locally on NVIDIA hardware or via cloud models. The local setup includes one-click inference and an optimized experience for NVIDIA hardware, utilizing the post-trained Qwen 3.8 27B model.

According to NVIDIA, the agent is designed to work optimally on its DGX Spark, giving customers flexibility to switch between local and cloud inference based on their needs. The local execution model allows developers to keep sensitive code, proprietary data, and project context within their own systems.

"When running locally, Perplexity offers one-click local inference setup and an optimized experience for local NVIDIA hardware — with a specially post-trained Qwen 3.8 27B model."

NVIDIA

NVIDIA also confirmed that Perplexity is working on a fine-tuned Nemotron 3.5 Lightning variant for ultra-fast responses, along with support for Windows and DGX Station.

The Portable Computer agent provides a local-first AI inference option that allows developers to maintain control of proprietary data while switching to cloud models when needed.

What Hardware Is Supported?

NVIDIA has made day-zero support for Portable Computer available on all NVIDIA RTX GPUs, including the RTX Pro series and the GeForce RTX 5090. The platform is optimized for NVIDIA DGX Spark, with additional support planned for DGX Station.

Developers can begin using the platform immediately with familiar open-source tools, all of which support Qwen3.8-27B on NVIDIA RTX GPUs. The combination of NVIDIA RTX performance, the CUDA software ecosystem, and these tools provides a foundation for local agent coding workflows.

Day-zero support for Portable Computer extends to all NVIDIA RTX GPUs, including the RTX Pro series and GeForce RTX 5090, with DGX Station support in development.

Who Is This For?

Portable Computer is designed for coders and developers who work with sensitive code, proprietary data, and project context that must remain secure within their own systems. The local execution model enables developers to polish work before sharing it with other developers.

The platform targets AI developers who require responsive local agent coding workflows without the recurring cost of cloud tokens. By offering both local and cloud inference options, the platform addresses developers who need flexibility between on-device processing and cloud-based models.

Portable Computer is intended for AI developers seeking local model execution to protect proprietary code and data while maintaining access to cloud inference options.

Common Questions

What is the difference between the Qwen 3.8 and PPLX local models?

Both the Qwen 3.8 and PPLX local models are capable of 27 billion parameters. The Qwen 3.8 model is post-trained for optimized local inference on NVIDIA hardware, while PPLX serves as an alternative local model option within the Portable Computer agent harness.

Which open-source tools support the Portable Computer agent?

Developers can use llama.cpp, Ollama, Unsloth, and LM Studio Bionic to get started immediately with Qwen3.8-27B on NVIDIA RTX GPUs. These tools are supported alongside the CUDA software ecosystem for local agent coding workflows.

When will the Nemotron 3.5 Lightning variant be available?

Perplexity is working on a fine-tuned Nemotron 3.5 Lightning variant for ultra-fast responses, which NVIDIA states will be coming soon. The exact release date has not been specified in the announcement.

Sources and Methodology

This article is based on the official NVIDIA announcement regarding the Portable Computer agent launch, as reported by Lowyat.NET. Direct quotes are attributed to NVIDIA and sourced from the NVIDIA blog post. The source URL is https://www.lowyat.net/2026/402525/nvidia-and-perplexity-launches-new-portable-computer-agent/. This article was last updated on June 5, 2026.