[joquarky]

And what hardware they needed to run the model, because that's the real pinch in local inference.