ProductJun 6, 2026, 08:00 PM
Huawei-led team claims it post-trained DeepSeek's 1.6-trillion-parameter model — 1,000 Ascend 910C chips used in training
Summary
A research group including Huawei Technologies says it completed full-parameter post-training of DeepSeek's V4-Pro, a 1.6-trillion-parameter model, using 1,000 Ascend 910C chips. This demonstrates the capability of Huawei's AI chips in large-scale model training and the adaptation progress of DeepSeek models on domestic computing platforms.
Why it matters
This event shows that Huawei's Ascend chips can support trillion-parameter level large model training, marking a milestone for the domestic AI computing ecosystem.
Source links
Market reaction
Huawei
Private / not listed
DeepSeek
Private / not listed