Hangzhou, China – May 20, 2026 – Alibaba Group has officially announced the launch of its next-generation AI accelerator, the Zhenwu M890, alongside the debut of its flagship large model, Qwen 3.7-Max. Unveiled at the 2026 Alibaba Cloud Summit, these releases mark a significant expansion of Alibaba's full-stack AI infrastructure, specifically designed to meet the surging demands of the "Agentic Era."
Developed by Alibaba’s Semiconductor subsidiary, T-Head, the new Zhenwu M890 is a high-performance training and inference chip. It boasts 144GB of mEMOry and features advanced interconnect Technology. To maximize its potential, Alibaba also introduced the Panjiu AL128 Supernode Server, a powerful system that links 128 Zhenwu M890 accelerators into a single rack, enabling massive concurrent inference and large model trAIning with ultra-low latency.
In a move to accelerate AI APPlication development, Alibaba simultaneously released Qwen 3.7-Max. This latest flagship model is optimized for complex coding and autonomous agent-based tasks, demonstrating Alibaba's commitment to shifting AI capabilities from simple responses to autonomous task execution.
Highlighting the rapid adoption of its hardware, T-Head revealed that the Zhenwu Series Has already shipped 560,000 units to over 400 external CLIents, including major enterprises like China Unicom. Looking ahead, the company confirmed a roadmap for annual upgrades to the Zhenwu series to sustain the fast-evolving AI landscape. These strategic advancements come as Alibaba explores a potential independent listing for its chip design business, T-Head.
Comments & Questions (0)
No comments yet
Be the first to comment!