Search
DeepSeek Opts for Huawei AI Chips in Model Training
1 week ago /
Read about 0 minute
Author：小编
DeepSeek has chosen to integrate Huawei's AI chips into its model training process, aiming to mitigate its dependence on NVIDIA chips. The newly introduced DeepSeek-V3.1 model features a hybrid inference architecture that seamlessly blends thinking and non-thinking modes. This innovative approach not only boosts model thinking efficiency and Agent capabilities but also refines tool utilization and enhances intelligent agent task execution. Furthermore, the model leverages the UE8MO FP8 Scale parameter precision, which is fully compatible with Huawei's Ascend chips, thereby bolstering both stability and efficiency.
-
C114 Communication Network -
Communication Home
7 X 24 Track global technological trends
