The 1% can afford to run these models without quantization.