HP Korea has domestically launched the 'ZGX Nano G1n AI Station,' an ultra-compact workstation designed for AI-related professionals, including developers and data scientists, on December 17, 2025. This device aims to address the limitations of existing desktop and laptop environments concerning large model processing, enhance the developer experience, and provide a foundation for enterprises to incrementally expand their AI workflows.

The ZGX Nano G1n AI Station is equipped with an advanced System-on-a-Chip (SoC) that integrates NVIDIA's Blackwell architecture-based 'GB10' GPU and an Arm IP-based 20-core Grace CPU. It features 128GB of LPDDR5X unified memory, enabling the execution of various AI tasks without reliance on cloud servers.

In terms of performance, the workstation delivers a maximum AI computation performance of 1,000 TOPS (Tera Operations Per Second). Its floating-point performance reaches approximately 1 PetaFLOPs (PFLOPS) under the FP4 (4-bit floating-point) standard. The system is capable of locally running AI models comprising up to 200 billion parameters. Furthermore, by connecting two ZGX Nano G1n units, the processing capability can be scaled to handle AI models with up to 400.5 billion parameters.

The software environment for the ZGX Nano G1n AI Station is based on Ubuntu Linux, specifically utilizing NVIDIA's DGX OS. It comes pre-loaded with the ZGX Toolkit, which integrates key development tools such as Visual Studio Code, MLflow, Ollama, and Unsolos. An extension for Visual Studio Code is provided to simplify development setup by automatically discovering and connecting to other ZGX Nano devices within the same network. For storage, users can choose between 1TB or 4TB NVMe M.2 SSD options, both supporting encryption. Purchase consultations and quotations are available through the HP Korea product website.