Quiz: Quantization and model size

What is the typical size reduction from FP32 to INT8 quantization?