And the fact that you will never be benefited from the batching they’ve explained either, you’ll be paying for every token sequential or parallel unlike nvidia or amd gpus