Alibaba Cloud's GP9A Instance Goes Live, Successfully Runs 2-Trillion-Parameter Model

By AI123

At the 2026 Apsara Conference on September 22, Alibaba Cloud announced that its Lingjun Zhenwu M890 supernode instance GP9A is now publicly available. The instance has become the first supernode-form computing service in China to successfully run a large model with over 2 trillion parameters.

A supernode typically refers to a server form that integrates multiple AI chips through high-speed interconnection into a large-scale computing cluster. The GP9A was showcased alongside other products such as Alibaba's Pingtouge Zhenwu V900 AI chip.

Alibaba also said a new supernode server powered by the Zhenwu V900 chip is expected to be available in the first quarter of 2027, and previewed its next-generation Zhenwu J900 AI chip, planned for a March 2028 release.

Explore topicsAlibabaCompute PowerModels

Source

View all