How We Measure
We run the exact same prompt on the same rig, changing only one variable per test.
The model is reloaded before every run.
Every prompt and its result file is published.
Hardware
One rig: NVIDIA RTX 3090 24 GB.
Model server: LM Studio.
About
PC Magik Lab explores running AI models locally on real GPUs.
Hosted by PC Magik, a computer service and homelab operator from Poland.