Report: DeepSeek Plans to Pretrain Next-Generation V5 Model on Huawei Ascend Chips
According to STAR Market Daily, DeepSeek plans to pretrain its next-generation V5 model on Huawei Ascend chips. Previously, Ascend only handled DeepSeek's inference and V4-Pro post-training tasks. X.PIN believes this is related to DeepSeek open-sourcing its complete Ascend compute and communication stack (DeepGEMM, FlashMLA, DeepEP) on September 30.
Why read this
SelectedAI score 73
The report comes from a media source; readers can use it to understand the possible shift in DeepSeek's training compute from Ascend inference to pretraining.
Topics
Body
DeepSeek plans to pre-train its next-gen V5 model on Huawei Ascend chips, per STAR Market Daily sources.
Until now, Ascend only handled DeepSeek inference and V4-Pro post-training. Suddenly, DeepSeek open-sourcing its full Ascend compute & comms stack (DeepGEMM, FlashMLA, DeepEP) on Sept 30 makes total sense.