NVIDIA and Perplexity have officially and jointly launched a new Agentic AI platform named simply as Portable Computer. Simply put, it’s an agent that doesn’t necessarily lean on the cloud, or the usual cost of tokens that run it.
As NVIDIA puts it, the Portable Computer agent is designed to work optimally on its DGX Spark, and offers customers the option to switch between running locally or via cloud models, allowing them to switch in between their needs.
“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. The company is also working on a fine-tuned Nemotron 3.5 Lightning variant for ultra-fast responses, which will be coming soon. Perplexity is also working on support for Windows, as well as DGX Station.“
Why This Matters, For AI Developers At Least
Portable Computer basically consists of an agent harness, an orchestrator, and the choice between the aforementioned Qwen 3.8 and PPLX local models. Both are capable of 27 billion parametres.
Because they’re local, coders and developers are able to keep sensitive code, proprietary data and project context all within their own system, away from other developers should they wish to polish it further before sharing.
In addition to DGX Spark, NVIDIA has also made day-zero support for Portable Computer available on all NVIDIA RTX GPUs, including the RTX Pro series and the GeForce RTX 5090.
“Developers can get started immediately with llama.cpp, Ollama , Unsloth and LM Studio Bionic, all of which support Qwen3.8-27B on NVIDIA RTX GPUs. Together, NVIDIA RTX performance, the CUDA software ecosystem and familiar open-source tools give developers broad support for new models and a strong foundation for responsive local agent coding workflows.”
(Source: NVIDIA)
