China Telecom Cloud is bringing in DeepSeek AI model onto its platform. It raises AI assistant for answering questions and such.
SiliconCloud platform is providing full industrial support for R1 and V3 using Ascend-910B GPUs
Full V3/R1 requires 16 GPU card and cost 90000 RMB/month
CAICT release first public service platform for computing power interconnection. In order to help domestic AI developers, it announced support for DeepSeek
Domestic GPU firm Sietium announced its self developed GPU is now compatible with running DeepSeek models. 1 GPU can run r! 32B model.
4-8 GPU can run R1 70B model.
Multiple machine cluster is needed to run V3 and full R1.
Nubia is working to integrate DeepSeek into Z70 Ultra's Nebula AIOS