Open Models Almanac
Methodsinference & servinginference quantization

single mention · inference & serving

NVFP4 routed-expert GEMM quantization

NVFP4 quantization for routed-expert GEMMs, with dynamic max-based activation scaling and max-calibrated 4/6 weight scaling.

The sources on this page only mention it in passing. The line above is the extraction's one-line gloss. See inference quantization for the methods filed beside it.

Also called NVFP4 quantization for routed-expert GEMMs, NVFP4 routed-expert GEMMs.

Filed under inference & serving › inference quantization