March 20, the latest ranking released by the authoritative large model blind test list LMArena shows that Alibaba's latest flagship preview model in the Qwen series, Qwen3.5-Max-Preview, performed strongly, scoring 1464 points and ranking among the top. This score not only broke the record for domestic models but also surpassed overseas top models such as GPT5.4, Claude4.5, and Grok4.1 in multiple dimensions. This update marks that China's self-developed large models have officially entered the global top tier in general capabilities, completely breaking the previous situation where a few overseas leading companies had long dominated the top positions.

QQ20260320-141020.jpg

In this blind test, Qwen3.5-Max-Preview demonstrated excellent logical reasoning and instruction following capabilities, with its score significantly outperforming domestic models such as Douluo 2.0, GLM5, and Kimi 2.5 at the same time. Meanwhile, LMArena also released the global ranking of comprehensive strength of large model companies, showing a cluster rise of Chinese AI power: among the top ten globally, five are Chinese companies, with Alibaba ranked in the top five globally and maintaining the top position in China, while ByteDance, Zhipu AI, Yuedao Dark Face, and Baidu also remain in the top ten.

Currently, the global AI competition has shifted from merely comparing parameter sizes to performance evolution based on real user preferences. Domestic models represented by Qwen are building stronger technological influence in the global developer community through rapid iteration and algorithm optimization. This transition from single-point breakthroughs to systematic leadership indicates that domestic models are reshaping the global AI computing power and model application competition landscape, providing solid technical support for the deep implementation of industry-level large models in the future.