Get Zlio on your home screen
One tap, full screen, like an app.
XDA explains how changing a single llama.cpp flag let an 8GB laptop GPU run a 35B-parameter model faster than the author can read, a practical trick for running large open models on consumer hardware.