yes, it has 8 exponent bits like float32 instead of 6 like float16