On August 12, news broke that Aliyun has officially launched its Lingjun Zhenwu M890 supernode instance, now available for purchase in the Ulanqab region. This marks a significant milestone as China's inaugural supernode-based computing power solution, specifically designed to support large-scale models boasting over 2 trillion parameters. Notably, Kimi K3 and Qwen3.8 Max are already leveraging this cutting-edge infrastructure to deliver their services. The Lingjun Zhenwu M890 stands out with its support for FP8 and FP4 low-precision computing, significantly expanding interconnection capabilities from 16 to 64 cards through the integration of the ICN Switch 1.0 chip. This advancement enables an impressive inter-card interconnection speed of 800GB/s. In training scenarios pertinent to intelligent driving and embodied intelligence, the M890 demonstrates a training performance that is threefold compared to its predecessor, the Zhenwu 810E.