阿里巴巴宣布推出 Qwen3.5，这是其旗舰 AI 模型的一次重大升级；当中国 AI 公司在 2025 年 R1 首次周年纪念前后、赶在 DeepSeek 下一个平台发布前加速竞逐时，该消息受到关注。新模型为 AI-agent 工作流程与多模态输入而打造，可处理文字、照片与影片，并宣称每段影片的分析能力最高可达 2 小时。

Qwen3.5 的到来，紧接在农历新年前几周内 ByteDance、智谱和 Minimax 一连串快速且具竞争力的发布之后；这段时期被视为中国消费者应用采用的关键，尤其是在为期 1 周的假期窗口期间。阿里巴巴的升级节奏也在加快：其于 2025 年 4 月推出 Qwen3 系列，并于上个月发布 Qwen3-Max-Thinking，以提升复杂推理与自主代理效能。

这场战略赌注同时涵盖技术与商业层面：执行长吴泳铭已承诺投入超过 $53 billion 于 AI 与基础设施，且随时间推进仍可能超越此水位；同时，阿里巴巴、腾讯与百度合计规划 4.5 billion yuan（$649 million）的节日聊天机器人激励。在消费端，通义千问正被整合进阿里巴巴的商业生态系，其 100 million 用户已可透过应用内代理完成例如点饮料与预订旅行等任务，显示出一条以规模为核心、指向留存与交易转化的推进路径。

On February 16, 2026 at 5:00 PM GMT+8, Alibaba announced Qwen3.5, a major upgrade to its flagship AI model, as Chinese AI firms race to launch before DeepSeek’s next platform near the first anniversary of R1 in 2025. The new model is built for AI-agent workflows and multimodal input, handling text, photos, and video, with a stated video analysis capacity of up to 2 hours per clip.

Qwen3.5 arrives after a rapid sequence of competitive releases from ByteDance, Zhipu, and Minimax in the weeks before Lunar New Year, a period viewed as critical for consumer app adoption in China during a 1-week holiday window. Alibaba’s upgrade cadence is also accelerating: it introduced the Qwen3 series in April 2025 and then launched Qwen3-Max-Thinking last month to improve complex reasoning and autonomous agent performance.

The strategic stakes are both technological and commercial: CEO Eddie Wu has committed more than $53 billion to AI and infrastructure, with potential to exceed that level over time, while Alibaba, Tencent, and Baidu together plan 4.5 billion yuan ($649 million) in holiday chatbot incentives. On the consumer side, Qwen is being integrated into Alibaba’s commerce ecosystem, and its 100 million users can already complete tasks like ordering drinks and booking travel through in-app agents, signaling a scale-focused push toward retention and transaction conversion.