On July 23, Alibaba Cloud made a significant announcement: its Zhenwu M890 super node has been successfully tailored to accommodate the Qwen3.8 large-scale model. This breakthrough is now accessible on the Alibaba Cloud Bailian platform, ready to deliver inference services with unparalleled efficiency.
Notably, this super node marks a pioneering achievement in China, being the first to support the operation of large-scale models boasting over 2 trillion parameters. The Qwen3.8 model, with its staggering 2.4 trillion parameters, finds a perfect match in the Zhenwu M890—a cutting-edge AI chip developed by T-Head. This chip seamlessly blends training and inference capabilities, accommodates multiple data precisions, and is versatile enough to cater to a wide array of application scenarios.
