Alibaba officially launched Qwen3.8-Max on August 3, 2026, positioning it as the most capable flagship model in the Qwen family to date. The release continues the lineage of the Qwen series and is built upon the architectural foundation of Qwen 3.5, scaled up to 2.4 trillion total parameters with 95 billion active parameters. This scale is intended to support end-to-end handling of complex tasks, producing reliable deliverables rather than just answering isolated questions. The announcement also noted that this is the first time a Qwen-Max-class model would have its weights open-sourced, with open weights slated to follow the initial launch.
Qwen3.8-Max is designed around broad improvements across coding, real-world work tasks, research, and long-horizon activities, with particular emphasis on taking multi-day coding projects from an empty folder to a finished result. The model natively incorporates visual intelligence, and both Alibaba's announcement and the Qwen team highlighted gains in arena evaluations, ranking fifth in Text Arena and second in Vision Arena. Practical access is provided through QwenCloud APIs for developers, and the same model powers Alibaba's QwenWork workplace agent platform. The combination of massive parameter count, multimodal capability, and long-context behavior makes Qwen3.8-Max well suited to enterprises and developers seeking a general-purpose assistant that can sustain complex, multi-step projects.
