Alibaba expands AI ecosystem with new V900 chip and Qwen models

Alibaba has announced a significant expansion of its artificial intelligence ecosystem, spanning hardware, mobile integration, and specialized generative models.

On the hardware front, Alibaba’s semiconductor arm, T-Head, unveiled the Zhenwu V900 AI accelerator. Designed for data centers, the chip reportedly offers three times the performance of its predecessor, the M890, and features 216 GB of memory. The company aims to build massive superclusters capable of scaling up to 500,000 chips to support frontier model training.

In the mobile sector, Alibaba is partnering with HONOR to integrate Qwen Intelligence into smartphones. This collaboration will enable devices like the Magic9 Series to use AI agents for task planning, cross-app execution, and creative image editing.

Alibaba also released several updates to its Qwen model family. The Qwen-Audio 3.1 series introduces advanced speech synthesis (TTS) with high controllability over emotion and tone, as well as real-time interaction capabilities. Additionally, the Qwen Image 2.1 model offers lightweight, 7B-parameter image generation and editing, featuring native support for transparent RGBA outputs. While highly capable on consumer hardware, the image model is currently restricted to non-commercial research use.