Alibaba's Qwen team has released Qwen DeepResearch 2511, a significant update to its AI-powered research assistant. The new version aims to lower the barrier to deep research, transforming it from a heavy task into a natural extension of thought.
The core philosophy behind DeepResearch is to eliminate the 'cost assessment' that often kills creative ideas before they are explored. By reducing the effort required to verify and develop a thought, the system encourages users to pursue curiosity-driven inquiries.
Key features include a dual-mode switch: Normal Mode for efficient general-purpose research, and Advanced Mode that invests more time and computing power for deeper multi-step reasoning. The system also supports local file uploads (PDF, Excel, images), automatically extracting and integrating content into the research process.
Enhanced web search and reading strategies reduce hallucinations and redundancy, while report writing is now more controllable, allowing users to adjust word count, paragraphs, and detail level. The interactive experience has been redesigned with clearer information hierarchy and intuitive citation marking. Additionally, outputs can be generated as web pages or podcasts with a single click.
According to Qwen, the new version (2511) shows comprehensive improvements over the previous 2507 version across multiple dimensions, including anti-hallucination rate, report comprehensiveness, reliability, instruction following, report length, search depth, and research time. These gains are driven by a multi-agent collaborative mechanism that optimizes user need analysis, strategy planning, search verification, parallel tool calls, web page reading, and report writing, combined with memory management and research state tracking.