Alibaba Qwen3.8-Max: 2.4 Trillion Parameters, Top AI Ranks

Alibaba has launched Qwen3.8-Max, its largest AI model to date. This new model has 2.4 trillion parameters and supports a context window of up to 1 million tokens. What's interesting is its performance: Alibaba states Qwen3.8-Max ranks fifth in Text Arena and second in Vision Arena. It handles tasks like coding, research, and multimodal processing. The model is available through APIs on Alibaba Cloud Model Studio, and its weights will be released next week. It can also be accessed via QwenWork. The architecture uses a Sparse Mixture-of-Experts design to balance scale and efficiency. While it has 2.4 trillion parameters, only 95 billion are active during operation, reducing computational cost. In internal tests, the model independently built a self-evolving agent framework over 16 days. It also outperformed human participants in a multimodal dialogue intent recognition challenge. The model is designed for diverse fields, from legal document review to architectural 3D modeling. It also supports visual intelligence, processing large volumes of visual or textual input. This allows it to perform tasks based on visual feedback, including editing raw video footage. The bottom line: This new model could significantly impact how various industries use AI for complex, long-duration tasks.