thinQmaster Virtual testing partner Adaptive practice exams that find your weak spots and drill them. Every answer explained. From $69
Enroll

thinQmaster

Find your exam

Five questions free, before you decide.

Step 1 · choose a vendor

In one of our camps? Your practice exams come with the tuition — they are on your study page already.

AI & Machine Learning · Intermediate

BF16

Brain Floating Point 16: a 16-bit format with FP32's exponent range but less mantissa precision.

Definition

BF16 keeps the dynamic range of FP32, so training is stable without loss scaling, making it the preferred 16-bit format for LLM training on NVIDIA Ampere and later GPUs and on Google TPUs.

Exam tip

BF16 vs FP16: same size, BF16 trades precision for range and is more stable for training.

Related terms