The Qwen team, known for its prolific AI model researchers and engineers at Chinese e-commerce giant Alibaba, has introduced significant enhancements to its Qwen Deep Research tool. This tool is available as an optional modality that users can activate in the web-based Qwen Chat, a ChatGPT competitor.
With this update, users will be able to generate not only comprehensive research reports with organized citations, but also interactive web pages and multi-speaker podcasts with just one or two clicks.
This feature is own releaseunlike many of Qwen’s previous open source model products.
This feature relies on an open source model, but Qwen3-Coder, kwen imagesand Quen 3-TTS To enhance its core functionality, it provides end-to-end experiences such as research execution, web deployment, and audio generation. Hosted and operated by Qwen.
This means users don’t need to configure any infrastructure and can benefit from managed, integrated workflows. That said, developers with access to the open source model could theoretically replicate similar functionality in private or commercial systems.
This update was announced today, October 21, 2025, through the team’s official X account (@Alibaba_Qwen) as follows:
“Qwen Deep Research has been significantly upgraded. In addition to reports, we now create live web pages and podcasts powered by Qwen3-Coder, Qwen-Image, and Qwen3-TTS. Insights are now both visual and audible.”
Multi-format research results
The core workflow begins with a user request within the Qwen Chat interface. From there, Qwen collaborates by asking clear questions to shape the scope of the investigation, pulls data from the web and official sources, and analyzes or resolves any discrepancies found. It also generates custom code if needed.
A demo video posted by Qwen on X uses Qwen Chat to explain this process using the US SaaS market as an example.
In it, Qwen pulls data from multiple industry sources, identifies discrepancies in market size estimates (e.g. $206 billion vs. $253 billion), and highlights ambiguity in the U.S. share of the global figure. The assistant comments on differences in scope between sources, calculates a compound annual growth rate (CAGR) of 19.8% from 2020 to 2023, and provides contextual analysis to support the raw numbers.
Once the survey is complete, the user can "eyeball" Click the icon below the output (see screenshot) to view the report in PDF format in the right pane.
Then, when viewing the report in the right pane, the user can "create" Click the button in the top right corner and choose from the following two options:
-
"web development" will generate Professional live web pageautomatically expanded, Hosted by Kwenuses Qwen3-Coder for structure and Qwen-Image for visuals.
-
"podcast," As stated, this produces audio podcastIt also features dynamic multi-speaker narration produced by Qwen3-TTS. Hosted by Kwen Easy to share and play.
This allows users to quickly convert a single research project into multiple formats of content (textual, visual, and audio) with minimal additional input.
This website contains inline graphics generated by Qwen Image and is suitable for use in public presentations, classrooms, or publications.
The podcast feature allows users to choose from 17 different speaker names as hosts and 7 as co-hosts, but I couldn’t find a way to preview the audio output before making a selection. It seems designed for deep listening on the go.
There was no way to change the language output displayed, so even though Qwen LLM is multimodal, my output was displayed in English, as was the report and the initial prompt. The voice was slightly more robotic than other AI tools I’ve used.
Here is an example of a web page I created about commonalities of historical authoritarian regimes, another web page about UFO or UAP sightings, and an example of a podcast about UFO or UAP sightings below this paragraph.
As you can see from our brief usage so far, the website is hosted via a public link, but the podcast must be downloaded by the user and cannot be linked publicly.
Please note that podcasts are very different from actual reports. Rather than just a read-aloud audio version, it’s a new format where the two hosts use the report as a starting point to discuss and joke about the subject.
The web page version of the report also includes new graphics not found in the PDF report.
Comparison with Google’s NotebookLM
While this new feature has been well-received by many early users, comparisons with other research assistants, especially Google, have also surfaced. notebook LMrecently exited beta.
Chubby (@kimmonismus), an AI commentator and newsletter writer, says this about X:
“I really appreciate that Qwen provides regular updates, which is great.
However, attempts to build a NotebookLM clone within Qwen-3-max look less promising compared to Google’s version. ”
While NotebookLM is built around organizing and querying existing documents and web pages, Qwen Deep Research has an even greater focus. Create new research content from scratchaggregates sources from the open web and presents it across multiple modalities.
This comparison suggests that although the two tools overlap in the general concept of AI-assisted research, they differ in their approaches and targeted user experiences.
availability
Qwen Deep Research is now open to the public. Qwen chat app. This feature can be accessed at the following URL:
As of this writing, pricing details for Qwen3-Max or specific Deep Research features have not been provided.
What’s next for Qwen Deep Research?
Qwen Deep Research aims to streamline the process from idea to publishable product by integrating research guidance, data analysis, and multi-format content creation into one tool.
The integration of code, visuals, and audio makes it especially appealing for content creators, educators, and independent analysts looking to extend their research into web- and podcast-friendly formats without switching platforms.
Still, it raises questions about how Qwen’s generalized approach stacks up in depth, precision, and sophistication when compared to more specialized products like NotebookLM. Whether the strengths of multi-format execution outweigh those concerns may depend on the user’s priorities and whether they value single-click publishing over tight integration with existing notes and materials.
For now, Kwenn shows that research doesn’t end with documents, it begins with them.
Let us know if you’d like to repackage this into something shorter, or tailor it to a specific audience, such as a newsletter, journalistic style blog, or internal team description.
