Testing Methodology
Consistency
Every test uses the same prompt to ensure fair comparisons across models.
Isolation
We test on the same rig, changing only one variable at a time per run.
Clean State
The model is reloaded completely before every single run to prevent memory artifacts.
Transparency
Every prompt and result file is published for full reproducibility.
The Rig
GPU
NVIDIA RTX 3090 (24 GB VRAM)
Model Server
LM Studio
About the Series
PC Magik Lab is a dedicated YouTube series exploring the performance and capabilities of local AI models on consumer hardware. We dive deep into how different models behave on real-world local GPUs, providing clear, reproducible data for the AI community.