On August 3, Alibaba officially released Qwen3.8-Max, the most powerful model in the Qwen family to date. Built on the Qwen3.5 architecture, the model expands to 2.4 trillion parameters with 95B activated parameters, supports 1 million token context, and covers full multimodal capabilities across text, image, audio and video.

In demonstrations, Qwen3.8-Max showcased long-horizon Agent capabilities such as 16 days of continuous autonomous coding without human intervention, with comprehensive improvements in coding, office and research scenarios, while inference prices dropped significantly, balancing performance and cost.

The industry generally believes that open-sourcing flagship models will further lower the barrier for enterprises to adopt large models, accelerating the rapid deployment of new application forms such as Agents in enterprise scenarios.