Chinese tech giant Alibaba introduced a new AI chip on Wednesday, marking the latest move by a domestic vendor to reduce its dependence on Nvidia GPUs. Unveiled at the Alibaba Cloud Summit, the Zhenwu M890 is a training-and-inference-integrated accelerator designed by T-Head, Alibaba’s specialized semiconductor unit.
According to Alibaba, the Zhenwu M890 is specifically architected for AI agents. The processor is optimized to handle the massive memory demands associated with long context windows and facilitates efficient communication between multiple AI models. The vendor debuted the chip alongside its updated Qwen 3.7-Max model, which is optimized to run on the M890 for up to 35 hours of continuous operation.
The Qwen 3.7-Max model demonstrates significant reasoning capabilities, supporting over 1,000 tool calls and sustained complex reasoning tasks. It is designed for sophisticated workloads such as multi-file code editing, refactoring, and rapid prototyping. Featuring a massive 1-million token context window, the model-chip synergy highlights Alibaba's push toward a full-stack AI strategy and technological self-reliance in the generative AI era.
