Detect browser hardware support and estimate what local AI models can run on your machine without blocking the first render.