What GPU do you actually need to run Llama 3 or Mistral on your own machine? A new tool called CanIRun.ai tries to answer that question before you waste an afternoon on failed installations.
The site works like the classic "Can You Run It" game compatibility checkers, but for AI models. You input your hardware specs - GPU model, VRAM, system RAM - and it tells you which open-source models your machine can handle, which ones will run slowly, and which ones are flat-out impossible. It covers popular models across different parameter sizes, so you can quickly see whether your RTX 3060 can manage a 7B parameter model (a model with 7 billion learned values - smaller and faster) versus a 70B one (much larger, needs serious hardware).
The timing makes sense. Local AI usage has grown sharply as models like Llama 3, Mistral, and Qwen have gotten good enough for real work. But the hardware requirements are confusing - VRAM matters more than RAM, quantized versions (compressed models that trade a bit of quality for much lower memory use) change the math entirely, and the difference between "runs" and "runs well" is significant. A quick lookup tool saves real time.
It is a simple utility solving a simple problem. Nothing fancy, but genuinely useful if you are evaluating whether to invest in a GPU upgrade or just want to know which models are realistic options for your current setup.