DeepSeek V4 Launches on Huawei Ascend Chips, Challenging OpenAI and Google
DeepSeek's V4 AI model, optimized for Huawei Ascend 950 chips, debuts with 1.6 trillion parameters and a 1M-token context window. The open-source framework cuts costs by 73% versus prior versions, outperforming rivals in reasoning and coding while sidestepping U.S. chip restrictions.
Collaboration with Huawei signals China's push for AI sovereignty. Technical leaps include hybrid attention architecture and 90% KV cache reduction—potentially reshaping global AI infrastructure dependencies.
Log in to Reply
Log in to comment your thoughtsComments
Related Articles
|Square
Get the BTCC app to start your crypto journey
Get started today Scan to join our 100M+ users