Does Model Optimizer support QAT for GLM-5 with FP8, INT4, and NVFP4 quantization?

How would you like to use ModelOpt

I would like to perform Quantization-Aware Training (QAT) on the GLM-5 model (744B parameters, MoE architecture) using Model-Optimizer, targeting the following low‑precision formats:

• FP8
• INT4 (full QAT, both weights and activations)
• NVFP4

Specifically, I want to verify official support and obtain any available examples or configuration templates for GLM-5 QAT with these precisions.

System information

CUDA version: 13
GPU: Blackwell