MSI Extends XpertStation WS300 and EdgeXpert Platforms with Support for NVIDIA Nemotron 3.5 Lightning and NeMo Switchyard

San Jose, California – Aug 11, 2026 – MSI today announced that its XpertStation WS300 and EdgeXpert platforms now support NVIDIA Nemotron 3.5 Lightning, a lightweight open model purpose-built for agentic AI and NeMo Switchyard, an open-source model routing library. Switchyard supports systems of models by routing each workflow step to the best-suited model, such as frontier-scale models for complex reasoning, lightweight models like Nemotron 3.5 Lightning for high-volume specialized tasks, improving accuracy, latency, and cost. With this support, XpertStation WS300, based on the NVIDIA DGX Station platform, serves as a deskside AI supercomputer for local AI development,, fine-tuning, and multi-model agent workflows, while EdgeXpert, powered by NVIDIA DGX Spark, applies the same routing intelligence in a compact system for always-on AI agents.

"No single system fits every stage of an AI agent's lifecycle," said Sam Chern, MSI Vice President of Marketing. "By supporting NVIDIA's latest routing and model technologies across both XpertStation WS300 and EdgeXpert, we're giving customers a complete path, from model training and fine-tuning to continuous agent deployment, without switching vendors or ecosystems.”

MSI XpertStation WS300 is built on the NVIDIA GB300 Grace Blackwell Ultra Desktop Superchip, with up to 748GB of coherent memory. This capacity allows Switchyard to run alongside multiple models locally, including both frontier-scale models and Nemotron 3.5 Lightning, without offloading any step to external infrastructure. Two WS300 units can be connected directly via an NVIDIA ConnectX-8 SuperNIC for larger workloads, while PCIe Gen5/Gen6 NVMe storage keeps data pipelines fed during sustained, multi-model workloads.

MSI EdgeXpert based on the NVIDIA DGX Spark platform, is powered by the NVIDIA GB10 Grace Blackwell Superchip, delivering up to 1 petaFLOP of AI compute with 128GB of coherent unified memory in a compact form factor. That footprint is enough to keep Nemotron 3.5 Lightning and models up to 200 billion parameters, running continuously on-device, so specialized, high-volume agent tasks stay local rather than routing out to the cloud. NVIDIA ConnectX-7 networking allows up to four EdgeXpert units to cluster together, extending Switchyard's routing across a larger pool of local models as agentic workloads grow.

NVIDIA NeMo Switchyard is an open-source model routing library for AI agents. It evaluates each workflow step and automatically selects the most suitable available model, with no training or manual configuration required, and improves its routing decisions the more agents run. NVIDIA Nemotron 3.5 Lightning is a lightweight open model distilled from NVIDIA's frontier Nemotron 3 Ultra. Built on a hybrid mixture-of-experts architecture with 30 billion total parameters and 3 billion active parameters, it delivers fast, efficient inference and supports context windows of up to 1 million tokens, making it well suited to run continuously alongside larger models as part of a complementary system of models.