Today (3rd), Alibaba unveiled a powerful artificial intelligence (AI) model called "Qwen 3.8-Max" with 2.4 trillion parameters, igniting competition for the lead in China's AI market. Following this news, Alibaba's stock price soared to 126 Hong Kong dollars during trading on the Hong Kong Stock Exchange, jumping 7.7% compared to the previous day.
'Qwen 3.8-Max' boasts a vast 2.4 trillion parameters, a scale comparable to 'Kimi K3' (2.8 trillion parameters) launched by Moonshot AI last month. The model adopts a 'Mixture of Experts (MoE)' design, using only 95 billion parameters per request, dramatically reducing operational costs and response latency. It also has the capability to handle various formats of data including text, images, and videos, processing up to 1 million tokens at once.
According to Alibaba's own announcement, 'Qwen 3.8-Max' ranked first in China for text models and second globally for visual analysis models on the AI system evaluation platform Arena. However, these rankings have not yet undergone official verification by independent benchmark institutions. With leading global AI players like OpenAI, Anthropic, and Google not disclosing their parameter scales, Alibaba's disclosure this time reveals the current state of China's AI technology development without reservation.
Particularly noteworthy is the fact that the 'Qwen' model is already being deployed in real-world applications and demonstrating its impact. The 'Qwen' model is powering Apple Intelligence's language-related AI features in China, which received approval from Chinese authorities on the 15th of last month. This suggests the practical market penetration capability and broad applicability of Alibaba's AI technology.