Alibaba Cloud Launches Lingjun Zhenwu M890 Ultra Node Instance
I'm LongbridgeAI, I can summarize articles.Alibaba Cloud launched the Lingjun Zhenwu M890 Ultra Node Instance in Ulanqab, enabling enterprise customers to access 64-GPU high-speed interconnected computing units for MoE large model inference up to 10T parameters. It is China's first supernode platform running models over 2T parameters, supporting Kimi K3 and Qwen3.8 Max. Featuring FP8/FP4 precision and ICN Switch 1.0, it triples training performance for autonomous driving and embodied intelligence compared to the previous generation.
Alibaba Cloud announced the official launch of the Lingjun Zhenwu M890 Ultra Node Instance, with the first batch now available in the Ulanqab region.
Enterprise customers can activate a 64-GPU high-speed interconnected computing unit on the cloud without building their own data centers, supporting inference for MoE large models with up to 10T parameters.
Alibaba Cloud's Lingjun Zhenwu M890 Ultra Node is also the first supernode-form computing platform in China to successfully run large models exceeding 2T parameters. Kimi K3 and Qwen3.8 Max are already providing external services through the instance.
The Lingjun Zhenwu M890 Ultra Node Instance supports FP8/FP4 low-precision computing. Through the ICN Switch 1.0 chip, Scale-up interconnection capacity has expanded from 16 GPUs to 64 GPUs, while inter-GPU connectivity has increased to 800GB/s.
In training scenarios such as autonomous driving and embodied intelligence, training performance has tripled compared with the previous-generation Zhenwu 810E.