Qwen4-Max released with million-token context and agent workflows
Qwen released Qwen4-Max, a multimodal model built on a new Qwen4-series architecture with 2.4 trillion backbone parameters, 71 billion active parameters per token, native visual capabilities, and a million-token context window. The model supports long-running agent workflows, professional tool use, and Live Delivery for iterative results, and is available through the Qwen platform, Token Plan, API, Qwen Studio, Qoder CN, and the Qwen App. Qwen also says it reduced API prices and open-sourced the Qwen-Live-Delivery-Plugin, while its benchmark, cost, and real-world performance figures remain vendor-reported.
Qwen4-Max adds a million-token multimodal model with long-running agent and live-delivery features to a major Chinese AI platform, making it relevant to practitioners building complex professional workflows.

我们正式发布 Qwen4-Max——与你并肩工作、创作和探索的全天候 AI 新同事。Qwen4-Max 基于 Qwen4 系列全新结构,拥有 2.4 万亿主干参数,具备原生视觉能力,支持百万 token 上下文窗口。
Qwen4-Max 以更强的专业工具运用、长程任务执行与成果交付能力,推动真实世界生产力的提升。从专业工作的深度到无限创意的广度,从 3D 建模与游戏开发到复杂精密的工业制造,从金融、法律、医疗到前沿科学探索——在与你的协作中,Qwen4-Max 能够理解工作全貌并持续推进项目,让每一个目标都拥有抵达现实的路径,实现人与 AI 协同工作的全新可能。

#Qwen4-Max#Qwen#Qwen-Live-Delivery-Plugin#Qwen Studio#Qoder CN#Alibaba Cloud#Work360#Terminal-Bench 4.0