Model techniques map
Techniquesoptimizationtraining precision

implementation detail · filed under optimization

BF16

The BF16 floating-point format is used for selected layers or precision-sensitive computations.

Also called BF16 precision, BF16 testing, keeping attention projections in BF16, BF16 (bfloat16).

sources
5
models
3
labs adopt it
2
strongest
used

How sources treat it

One count per evidence span, weakest treatment to strongest.

evaluated 1used 4

Documented in

Further reading

Picked by hand, not extracted: where to read more, not evidence for anything on this page.

Evidence

5 spans quoted from the sources, strongest treatment first.

Released under Apache 2.0 with BF16, FP8, NVFP4, and GGUF weights on Hugging Face

usedsoftware implementationin Step 3.7 FlashStepFun

BF16 precision for attention linear layers, MoE latent projection layers

usedpost trainingin Nemotron 3 UltraNVIDIA

select layers (including latent projections, MTP layers, QKV/attention projections, and embeddings) are maintained in BF16 or MXFP8 for training stability

usedmodel architecturein Nemotron 3 UltraNVIDIA

we kept the QKV and attention projections in BF16

usedoptimizationin Nemotron 3 familyNVIDIA

our tests show minimal intelligence impact compared to BF16 testing

evaluatedevaluation onlyin Nemotron 3 UltraNVIDIA

Filed alongside

Other methods under optimization :: training precision.