Local models, real GPUs.
Every test follows the same strict protocol — so the only thing that changes is what matters.
Every model is tested with the exact same prompt, ensuring a fair apples-to-apples comparison.
All benchmarks run on the same hardware — one GPU, one setup, no variables except the model itself.
Only one parameter changes per test — model, quantization, or context — keeping results clean and interpretable.
The model is fully reloaded before every single run, eliminating caching or state carryover.
Every prompt, every result file, every configuration — all published openly for full transparency.
One machine. One GPU. Consistent conditions for every benchmark.
Every test runs on this exact configuration
PC Magik Lab is a YouTube series dedicated to running AI models locally on real, consumer-grade GPUs. No cloud APIs, no abstract benchmarks — just real models, real hardware, and real results.
Each episode puts a model through a controlled test on a single NVIDIA RTX 3090 rig, changing only one variable at a time. Every prompt and every output file is published for full transparency.
The goal is simple: show what AI models actually do when you run them on your own machine — and what it really takes to get them running.
Watch the episodes, inspect the data, or both.