Two Qwen3 models on one DGX Spark: the residency math

The author describes their experience setting up a local LLM (Large Language Model) on a DGX Spark, discussing the residency math, GPU memory utilization, and the importance of empirical sizing.