2 ms·

Then there's standard formats like ONNX and inference on any platform or language or hardware that you prefer