Deepseek R1 and Alibaba’s Qwen 2.5 have now been deployed on Phoenix and are ready for inference via PhoenixLLM.

More importantly they are running on our new Huawei Ascend High Performance Cluster on Phoenix SkyNet, which initially consists of 200+ nodes sourced from various partners such as China Telecom Global and Tensor (https://t.co/iNeAL2D5JY). These nodes house a combination of the latest Huawei Ascend 910C & 910B GPUs – of which 910C delivers 60% performance vs an Nvidia H100 for inference, and similar or better performance than Nvidia A100.

We are in the process of porting various AI applications and models from Phoenix GenAI, AlphaNet, and open source AI over to support the Huawei Ascend architecture for both inference and training. This is only the beginning for Phoenix’s AI & compute DePIN expansion to provide and integrate powerful alternatives to mainstream computing paradigm.

$PHB #Deepseek #ChinaAI #HuaweiAscend