Our team walks you through how the NVIDIA DGX Spark performs in common AI development scenarios. In this video, we demonstrate local LLM inference and fine-tuning on the 128GB unified memory, show how two units link over ConnectX-7 for 405B-parameter scale, and explain what 1 PFLOPS FP4 means in practice. You'll see the DGX OS environment, model loading, and a walkthrough of the ports and setup — real use insights for B2B teams evaluating desktop AI workstations.