Open Models Almanac
Methodsinference & servinginference quantization

implementation detail · filed under inference & serving

FP8

The FP8 precision format, mentioned as a model-weight format without a more specific quantization procedure.

The sources on this page name it without explaining how it works. The line above is the extraction's one-line gloss. See inference quantization for the methods filed beside it.

Also called FP8 (8-bit floating point), FP8 quantization, --quantization fp8, FP8 model, FP8 quantized instruct model, Hy4 preview-FP8.

sources
18
models
12
labs adopt it
7
strongest
default

Documented in

Evidence

18 quotes from 18 sources, one card per model, each opening with its strongest.

How sources treat it
optional 5used 11default 2

Hy3

Tencent (Hunyuan)adopted2 sources

A separate Hy3-FP8 checkpoint is also released. FP8 lowers the memory footprint for cheaper serving.

optionalpost trainingin Hy3Tencent
1 more from this model

Hy3-FP8 | FP8 quantized instruct model

usedsoftware implementationin Hy3Tencent Hy Team
↑