NVIDIA said it has achieved a record for large language model inference, announcing that an NVIDIA DGX B200 node with eight Blackwell GPUs achieved more than 1,000 tokens ….