参考资料 6可信度 87%总体可信度: 87%该图钉的来源和参考资料对其日期的支持程度。包括来源在内的 6 份资料对图钉开始和结束时间支持程度的加权平均；资料每比最新的一份旧 180 天，权重减半显示所有可信度不低于 75% 的图钉
第一项始终是图钉的来源。总体可信度是各份资料对上方所用开始和结束时间支持程度的加权平均；资料每比最新的一份旧 180 天，权重减半。
- [1]90%qwen.ai/blog?id=qwen3.6qwen.ai· 发布于 2026年10月12日· 开始 2026年4月2日· 占评分 26%
财新全球,2026年4月2日:“Alibaba Group Holding Ltd. on Thursday released Qwen 3.6-Plus”(4月2日);Model Studio列出了发布时快照qwen3.6-plus-2026-04-02,OpenRouter[6]显示“Released Apr 2, 2026”。Qwen博客页面标注日期为2026/04/01,阿里云社区转载于4月2日。
- [2]90%qwen3.6-plus Model Info - Alibaba Cloud Model Studiohelp.aliyun.com· 添加于 2026年10月12日· 占评分 26%
Model Studio page: native vision-language Plus model with image, text and video input and text output, markedly better agentic coding, front-end and vibe coding than the 3.5 series; snapshot qwen3.6-plus-2026-04-02; function calling, structured outputs, web search and context caching; CNY 2 input and CNY 12 output per million tokens up to 256K input (CNY 8 and CNY 48 up to 1M).[1]
- [3]80%Alibaba Unveils Qwen3.6-Plus to Advance the Agentic AI - Alibaba Cloud Communityalibabacloud.com· 发表于 2026年4月3日· 占评分 12%
Alibaba's release: the model will be integrated into Wukong, its AI-native enterprise platform, and the Qwen[1] app; strengths in visual coding (screenshots and wireframes to front-end code), long-form video reasoning and repository-level engineering.
- [4]92%Alibaba Releases Qwen 3.6-Plus AI Model With Enhanced Coding Capabilities - Caixin Globalcaixinglobal.com· 发表于 2026年4月2日· 开始 2026年4月2日 ✓· 占评分 12%
Caixin, 2 April 2026: Alibaba "on Thursday released Qwen[1] 3.6-Plus", pitched on agentic coding that breaks down tasks, writes and tests code and troubleshoots iteratively; native multimodal; available on Alibaba Cloud's[3][5] Bailian platform at CNY 2 per million input and CNY 12 per million output tokens; part of an effort to regain momentum in the AI race.
- [5]85%Qwen3.6-Plus: Towards Real World Agents - Alibaba Cloud Communityalibabacloud.com· 发表于 2026年4月2日· 占评分 12%
Repost of Qwen's[1] launch post: available immediately via API after the Qwen3.5 series in February, hosted on Model Studio with a 1M-token context by default, better agentic coding and multimodal perception, and built in response to community feedback on Qwen3.5-Plus; benchmark table against Qwen3.5-397B-A17B, Kimi K2.5, GLM5, Claude 4.5 Opus and Gemini 3 Pro.