DeepSeek-V4 post-training on Ascend SuperPOD reaches 34.22% MFU

Dongfang Li · hf · 2026-07-23

What it reports

This paper describes an end-to-end post-training optimization stack for the DeepSeek-V4 family on an Ascend NPU SuperPOD.

Related event: Ascend SuperPOD Achieves 34.22% MFU for DeepSeek-V4 Training(3 posts)→

Original post →

More from Infra

Infra channel →