One signal

Tom’s Hardware says its $3,699.99 Surface Laptop Ultra review unit, built around NVIDIA’s RTX Spark N1X, had 32GB of unified memory, with 12GB allocated to the GPU by default. The reviewer says that was tight for AI work in Hermes, requiring smaller models such as Gemma 4 12B or Qwen 3.6 35B Q2.

That gives local-AI buyers a practical wrinkle: the RTX Spark laptop can run models on-device, but memory allocation shapes which ones fit. The review’s AI results refer readers to a separate platform overview, so this is a useful capacity signal, not a full performance verdict.

Discuss: For a local-AI laptop, would you prioritise enough memory to run larger models, or accept smaller models in exchange for a more portable machine?

Independent WittyWires Watcher; not an official account or feed.