Ali (09988) Qwen3.5-Max-Preview stunning debut, topping the LM Arena China charts.

Alibaba's flagship model Qwen3.5-Max-Preview has made its official debut on the global large model competition platform LM Arena, receiving a comprehensive score of 1464 points. This not only breaks the record for domestic models, but also signifies China's official entry into the top five in the global AI competition for computing-intensive models. It is rumored that the official version of Qwen3.5-Max will be released soon, with its capabilities continuing to evolve and upgrade. LM Arena is considered the most practical large model ranking list in the industry due to its use of the "black box battle + global developer voting" mechanism. The listing of Qwen3.5-Max-Preview this time may indicate that domestic large models have officially entered the top global tier in terms of general capabilities, logical reasoning, and multilingual comprehension. Qwen3.5 is the latest generation of large models released by Alibaba on the eve of the Chinese New Year this year. Currently, 8 different models ranging from 0.8B to 397B in size have been open-sourced, with a relatively complete open-source gradient. Among them, Qwen3.5-Plus, with around 397 billion total parameters and 17 billion activation parameters, stands out among models of the same size. The Qwen3.5-Max-Preview that made its appearance this time is the preview version of the flagship model in this series, ranking first in China in terms of overall performance, expert-level text capabilities, and mathematical abilities. In addition, LM Arena ranks global large model labs based on the strongest models from each company. Five Chinese companies have entered the top 10, with Alibaba ranking in the top five globally and first in China. ByteDance, KNOWLEDGE ATLAS, Dark Matter of the Moon, and Baidu are also among the top 10 globally.