Alibaba Unveils Qwen4-Max: Its Most Powerful AI Model Yet Challenges Global Leaders
Alibaba has released Qwen4-Max, the new flagship model in its Qwen series and the company’s most capable AI system to date. Announced on November 16, 2026, the model delivers stronger performance on complex professional and scientific tasks while cutting costs and tightening safety controls. It positions Alibaba more firmly in the intensifying race against U.S. leaders such as Anthropic and OpenAI, and against domestic rivals including Moonshot AI. Qwen4-Max is a multimodal mixture-of-experts (MoE) model with 2.4 trillion total parameters. It activates fewer parameters per token than its predecessor (roughly 71 billion versus about 95 billion in Qwen3.8-Max), supports a one-million-token context window, and handles text, images, and video natively. Alibaba positions it as an “AI teammate” for creative work, professional workflows, and frontier research. Key Technical and Performance Highlights Qwen4-Max improves on Qwen3.8-Max (released in August 2026) by achieving higher results ...