https://en.wikipedia.org/wiki/Bfloat16_floating-point_format

Yes, however it’s a different format from standard fp16, it trades precision for greater dynamic range.