Search
On March 20, LMArena, a globally recognized blind-testing ranking platform for large language models, released an updated set of rankings. Alibaba’s latest flagship model preview version, Qwen3.5-Max-Preview, made its debut in the competition and achieved a remarkable score of 1,464 points, outperforming leading overseas models such as GPT-5.4 and Grok-4.1.
Qwen3.5 represents Alibaba’s newest series of large language models, launched on the eve of this year’s Chinese New Year. The series encompasses eight models of varying sizes, ranging from 0.8 billion to 397 billion parameters, all of which are open-sourced. The Qwen3.5-Max-Preview, evaluated as the flagship preview version in this round, ranks first in China in terms of overall performance, expert-level text generation capabilities, and mathematical reasoning abilities.
-
C114 Communication Network -
Communication Home
