Local models, real GPUs.
Every test uses the exact same prompt and the exact same hardware rig to ensure absolute comparability.
We change only one variable per test, removing guesswork and pinpointing exactly what affects performance.
The model is fully reloaded before every single run to eliminate caching bias and memory fragmentation.
Every prompt used and every resulting output file is published openly for community verification.
PC Magik Lab is dedicated to exploring the capabilities of Large Language Models running on consumer-grade hardware. By focusing on local execution and rigorous testing, we provide real-world insights into how different models and configurations perform on real GPUs.