Alibaba unveils Qwen3.7 AI models and Zhenwu M890 chip with 3x performance

At the Alibaba Cloud Summit, the Chinese tech giant unveiled preview versions of its Qwen3.7 AI models and officially launched the Zhenwu M890, a next-generation AI processor. The Qwen3.7-Max-Preview model now ranks 13th globally for text-based tasks, while the Qwen3.7-Plus-Preview achieves 16th place for vision on the Arena leaderboard, demonstrating significant progress in both language and multimodal capabilities. These models are designed for cloud-native deployment, targeting enterprise customers who require high-performance inference and training on Alibaba Cloud infrastructure.

Meanwhile, the Zhenwu M890 chip marks a major leap in Alibaba's in-house silicon efforts. It delivers three times the performance of its predecessor, the Zhenwu 810E, and features 144 GB of GPU memory, making it suitable for memory-intensive AI workloads like large language model training and real-time inference. This chip is part of Alibaba's strategy to reduce reliance on foreign suppliers and power its cloud services with proprietary hardware. Combined, the Qwen3.7 models and Zhenwu M890 position Alibaba as a formidable competitor to Western AI leaders like Nvidia, OpenAI, and Google in both software and hardware.