The Alibaba Zhenwu M890 supernode has successfully adapted to Qwen 3.8 and launched the Alibaba Cloud Bailian platform to provide model inference services. According to Alibaba Cloud's official report, this is the first domestic supernode to successfully run a model with large parameters exceeding 2 trillion yuan. Qwen3.8 is Ali's latest flagship model. The parameter scale is as high as 2.4 trillion yuan. The model that runs through this scale needs to distribute the parameters to dozens or even hundreds of high-speed connected GPU cards, so that the trillion parameter “runs steadily and fast.” However, due to limited communication bandwidth, ordinary AI clusters cannot meet the requirements of high throughput, low latency, and high concurrency.

The Alibaba Zhenwu M890 supernode has successfully adapted to Qwen 3.8 and launched the Alibaba Cloud Bailian platform to provide model inference services. According to Alibaba Cloud's official report, this is the first domestic supernode to successfully run a model with large parameters exceeding 2 trillion yuan. Qwen3.8 is Ali's latest flagship model. The parameter scale is as high as 2.4 trillion yuan. The model that runs through this scale needs to distribute the parameters to dozens or even hundreds of high-speed connected GPU cards, so that the trillion parameter “runs steadily and fast.” However, due to limited communication bandwidth, ordinary AI clusters cannot meet the requirements of high throughput, low latency, and high concurrency.