User shares eGPU setup for local inference with Qwen3.6 via llama.cpp

A Reddit user describes using an eGPU on a laptop to run Qwen3.6 35B A3B with llama.cpp. The post highlights llama.cpp's built-in benchmark feature for tuning inference on modest hardware.