Feb. 20 (Bloomberg) -- According to Sina Tech.AliAlibaba Group CEO Wu Yongming said at the Q3 FY2025 earnings conference that Ali will soon release a Qwen2.5-MAX based on the Qianwendeep inference model.
1AI notes that on January 29th, on the occasion of the New Year, AliCloud announced its new Tongyi Thousand Questions Qwen 2.5-Max hyperscale MoE model, which is claimed to have been tested in a number of benchmarksBeyond DeepSeek V3 and other competitors.
AliCloud then announced that Qwen2.5-Max outperformed models such as DeepSeek-V3, Open AI o1-mini and Claude-3.5-Sonnet in the Chatbot Arena large model blind test, and ranked seventh in the world with a score of 1332.Also the non-reasoning Chinese Grand Modeling Champion. Meanwhile, Qwen2.5-Max ranked first in individual competencies such as Math and Programming, and second in Hard prompts (Hard prompts).