NVIDIA And Perplexity Launch New “Portable Computer” Agent

NVIDIA and Perplexity have introduced a new AI platform called Portable Computer. This agent operates independently from the cloud, eliminating the usual token costs associated with running AI models. The Portable Computer is specifically designed to function optimally on NVIDIA's DGX Spark hardware, offering users the flexibility to run the agent either locally or via cloud-based models, depending on their requirements.

The Portable Computer consists of three main components: an agent harness, an orchestrator, and a choice between two local models - Qwen 3.8 27B and PPLX. Both models boast an impressive 27 billion parameters, ensuring powerful performance on the local hardware. NVIDIA is also developing a fine-tuned Nemotron 3.5 Lightning variant, which is expected to be released soon.

The advantages of using Portable Computer are particularly significant for AI developers. By operating locally, developers can securely store sensitive code, proprietary data, and project context within their own systems, away from potential exposure to other developers. This level of control allows for further refinement and polishing of the AI models before sharing.

Portable Computer is compatible with NVIDIA RTX GPUs, including the RTX Pro series and the GeForce RTX 5090. Developers can begin using the platform immediately with llama.cpp, Ollama, Unsloth, and LM Studio Bionic, all of which support the Qwen3.8-27B model on NVIDIA RTX GPUs. The combination of NVIDIA's RTX performance, the CUDA software ecosystem, and familiar open-source tools provides developers with a robust foundation for creating responsive local agent workflows.