Bfloat16

Brain Floating Point 16-bit format, a truncated version of the 32-bit float format designed to accelerate machine learning computations and save memory.

Part of the Data & AI Terms glossary.