🤖 Alibaba releases Qwen4-Max: 2.4 trillion parameters and 62% reduction in task costs
On November 16, 2026, Alibaba introduced the Qwen4-Max flagship based on the Qwen4 architecture: a 2.4 trillion parameter MoE backbone (71 billion active per token) plus 328 billion N-gram Embedding, with a 1 million token context and input support for text, images, and video.
🌍 According to Qwen, the model scored 52.69 on the Work360 benchmark compared to 47.74 for Qwen3.8-Max, while the average task cost dropped from $4.43 to $1.67. Cheap long context reduces the cost of agent scenarios and increases price pressure on competitors.
👤 The API is available in QwenCloud, Qwen Studio, QwenWork, and Qoder: $1.50/$4.50 per million tokens, with caching at $0.075. These figures are vendor self-reported and can be verified immediately on your own tasks; weights are closed, but smaller Qwen4 models are promised to be open-sourced.
Source 1: https://qwen.ai/blog?id=qwen4-max Source 2: https://finance.biggo.com/news/357f82fc-873e-4025-8c0e-f2b920922b2b
