Test: 32GB RTX Spark loses to Snapdragon X2 laptop at running local AI models

A comparison of the $3,200 HP OmniBook Ultra 16 with a 32GB RTX Spark and the $2,200 ASUS Zenbook A16 with a Snapdragon X2 Elite Extreme and 48GB of RAM found that Hermes loaded no models on the RTX Spark, while the Zenbook loaded three in one click. On the RTX Spark, Qwen 3.6 35B-A3B ran only after manual tweaking in LM Studio and crashed often.

• The $2,200 Zenbook A16 loaded Qwen 3.8 27B, Nemotron 3.5 Bolt and Qwen 3.6 35B-A3B in one click
• With 16GB of dedicated VRAM, the RTX Spark found no compatible models in Hermes
• On the Zenbook the model ran on CPU due to Adreno GPU driver errors and took minutes per prompt
• A Python script with pytest prompt on the Zenbook did not finish within an hour