QWEN • PRESENTATION GUIDE
Qwen Slides: create a presentation and check the download
To make a presentation in Qwen, open the official assistant, choose Slides from Select Mode, describe the audience and source material, then review the proposed outline before confirming generation.
Our recorded results: the September 25 first deck changed two attendance figures and used misleading charts. The September 26 follow-up applied supplied corrections to the counts and financial table, but its attendance chart still misrepresented the values. Both downloads were six-page PDFs made of slide images; we did not receive a native editable PowerPoint file.
By China AI Chat Team · First tested September 25, 2026 · Correction trial September 26, 2026 · Mode menu checked September 30, 2026 · How we research. ChinaAI is independent of Qwen and Alibaba.
How to use Qwen Slides
- Open the official assistant at chat.qwen.ai. Use a fresh conversation for a new presentation and sign in if prompted.
- Open the plus menu (Select Mode) and choose Slides. The menu checked on September 30, 2026 still listed Slides. In our original trial, selecting it showed a Slides label beside the composer.
- Provide a clear brief. State the audience, purpose, slide count including the cover, source facts, style and any required calculations. We pasted a fictional source packet into the prompt.
- Review the draft plan before confirming. Check whether it covers your decision, keeps the requested order and reserves room for the conclusion. Ask for a specific adjustment if it misses something.
- Confirm the plan when it is ready. Qwen asked us to reply “Confirm” or “Start” to generate the full deck. For our first-result test, we replied only “Confirm” and preserved its initial plan unchanged.
- Inspect every rendered slide and download the available file. Check the exported format and content separately. An attractive preview or the word “PPT” does not establish that you have an editable PowerPoint file.
The first outline changed the requested sequence and lacked a clearly named final decision slide. We kept it for the first-result test. In your own deck, resolve outline omissions before generating; our later correction trial began from the completed presentation.
Give Qwen a source brief it can work from
Separate what the audience needs to decide from the visual style. Visual polish cannot compensate for missing data or an unclear purpose. Name the figures that must remain unchanged, the denominator for each percentage and the information that is not available.
This is a suggested prompt template, not a separately tested request:
Create [number] slides, including the cover, for [audience] to [understand or decide a specific issue]. Use only the source material below. Include [required sections] in this order. Keep [dates, values, units and definitions] unchanged. Mark missing information rather than inventing it. Use [readable style and frame shape]. Show [specific calculation or chart] with labelled values. Finish with [the decision or next action]. Provide the actual available download format and explain whether its text and charts are editable. Source: [your material].
Only provide material you are allowed to send to an online service. For this guide, all names, attendance figures and costs are fictional. No private museum records were uploaded.
Our test: a museum deciding whether to extend late opening
We requested six slides for the manager of fictional Northbank Museum. Four Thursday evenings in September 2026 had capacity for 80 visitors each. Attendance was 40, 55, 50 and 75. Each admission cost $8, staffing cost $120 per evening and materials cost $2 per attendee. Marketing expenditure and satisfaction had not been measured.
The brief asked for a labelled attendance chart, a cost calculation and a conditional decision. It explicitly prohibited invented causes, forecasts and a net-profit claim. We calculated the expected answers before submission and kept the first presentation. Download the exact prompt, source checks and evidence.
What our first Qwen presentation got right and wrong
On September 25, 2026, we sent one brief, confirmed the first outline and kept the first completed presentation. We inspected all six downloaded PDF pages as well as the app preview. See the test conditions for the displayed model, mode and limits.
The useful part: the deck retained the overall $1,760 revenue, $920 listed costs and $840 contribution, included a “Not net profit” warning, and asked for marketing-cost and satisfaction checks. Its large headings and main text were readable in the reviewed export.
The blocking problems: the attendance chart changed the middle two weeks, its capacity line contradicted the values, and the financial charts made $840 look larger than $920. It also omitted the total attendance, overall utilization and separate staffing/materials calculations we requested.
For a manager, these are decision-relevant errors. The first deck should be corrected and checked again before presenting it as evidence.
| Evening | Attendance | Capacity | Utilization |
|---|---|---|---|
| Week 1 | 40 | 80 | 50% |
| Week 2 | 55 | 80 | 68.75% |
| Week 3 | 50 | 80 | 62.5% |
| Week 4 | 75 | 80 | 93.75% |
| Total / overall | 220 visits | 320 places | 68.75% |
| What we checked | What the first deck showed | Practical implication |
|---|---|---|
| Six slides and requested structure | Six pages, but the sequence changed. September 2026 was absent from the rendered slides. | Review the outline for missing sections, then check dates again in the final pages. |
| Attendance and utilization | Page 3 showed only Weeks 1 and 4, with 50% and a reasonably rounded 94%. Page 4 changed Weeks 2 and 3 to 52 and 63. The 220-visit total and 68.75% overall utilization were missing. | Correct first and last values do not prove the intermediate weeks or overall rate are correct. |
| Attendance chart | The 80-person capacity line was below the bars labelled 52, 63 and 75. | Check axis and reference-line geometry, not only printed labels. |
| Financial calculation | The $1,760 − $920 = $840 subtraction was correct. Separate $480 staffing and $440 materials calculations were absent. | Keep the components visible so another person can audit the total. |
| Financial charts | On pages 5 and 6, the $840 contribution bar was taller than the $920 listed-cost bar. | The written figures were right, but their visual comparison was wrong. |
| Uncertainty and decision | Marketing and satisfaction checks remained, with a “Not net profit” warning and “Verify Before Extending.” The ending referred to a board decision rather than the requested manager. | The caution was useful, but the audience and final decision still needed editing. |
| Readability and presentation shape | Main text was readable without obvious clipping. The exported pages were approximately 7:4, rather than the requested 16:9. | Check the actual page shape and fit in the software you intend to present with. |
These findings describe the first result from one fictional brief. We subsequently tested one assisted correction. Revising the initial outline before first generation, other templates and repeated reliability remain untested. The original PDF and full check record let you inspect the result yourself.
What happened when we asked Qwen to correct the deck?
On September 26, 2026, we asked Qwen to correct the completed presentation, then confirmed its new outline. It produced a separate six-slide deck; the original remained accessible.
This was assisted correction. We supplied the correct counts, percentages, totals and financial calculations. We also asked for a financial table instead of the misleading financial bars. This checks whether Qwen applied explicit corrections; it does not test independent arithmetic or prove that it repaired the original financial charts.
The revised deck still needed work: attendance labels and the financial table improved, but the attendance chart remained misleading, requested percentages were missing, and export shape and editability were unchanged.
| Check | September 25 first deck | September 26 revised deck |
|---|---|---|
| Counts and totals | The chart changed 55 and 50 to 52 and 63; aggregate attendance and capacity were missing. | Page 3 kept 40, 55, 50 and 75, all four denominators of 80, and 220/320. It correctly distinguished visits from unique people. |
| Percentages | Only first/last-week rates appeared; the overall rate was absent. | The four requested weekly percentages were absent. “68.75% capacity” appeared above the chart without clearly identifying it as the overall 220/320 rate. |
| Attendance chart | An 80-person reference sat below three bar tops. | The bar labelled 50 was shorter than the bar labelled 40. There was no labelled zero, numeric scale or clearly identified 80-person capacity line. The dashed line did not establish the requested reference. |
| Financial calculation | Correct overall subtraction, missing $480/$440 components and misleading bars. | Page 4 showed the correct $1,760 revenue, $480 staffing, $440 materials, $920 listed costs and $840 contribution with calculations and a not-net-profit warning. The requested table replaced the bars. |
| Structure and decision | Date missing; final audience shifted to a board. | September 2026 and fictional-data labels appeared. Page 2 remained a narrative introduction rather than the requested source table. The ending kept a conditional decision but did not explicitly name the manager or one-month extension. |
| Remaining ambiguity | Material numeric and chart errors. | Page 5 labelled $840 “unverified”, although the listed-cost calculation was correct; full profitability remained unknown. Page 6 used a confusing decision diagram with a Yes arrow entering the diamond and two No labels. |
| File and readability | Readable six-page image PDF, approximately 7:4. | Readable main content in another six-page image PDF, still approximately 7:4 rather than 16:9. No native editable PPTX was received. |
The practical lesson from this one follow-up is to recheck the rendered output after a correction, including parts that already looked right. Naming the exact errors helped some content, but did not make the presentation ready to use. We kept the first revised output and did not request another generation to improve the result.
Can Qwen Slides export an editable PowerPoint?
In our September 25 and 26 tests, both downloads were PDFs, not native editable PPTX files. We used the presentation preview’s Download menu. No native PowerPoint export or slide-object editor was visible in the inspected workflow.
Each PDF contains six slide pictures, without separate editable text boxes or chart data. We checked the actual files, not just their names. Both have an approximately 7:4 page shape, rather than the requested 16:9. The file inspection details record the dimensions and structure.
Choose the download you want to inspect
Both PDFs preserve Qwen’s outputs and their remaining errors. The revised file followed a request that supplied the correct values; it is not a replacement for the original record.
On a narrow screen, scroll the table sideways to see every column.
| Presentation | Test date | PDF size | Main problems still in the file | Request and check records |
|---|---|---|---|---|
| Download original first PDF | 25 September 2026 | 32.1 MB | Changed attendance figures, misleading attendance and financial charts, and missing totals and cost components. See the first-result checks. | Download prompt and check records (ZIP) |
| Download revised PDF | 26 September 2026 | 30.9 MB | Counts and the financial table improved, but attendance-chart geometry, missing percentages and decision ambiguities remained. See the correction checks. | Download correction request and checks (ZIP) |
Sizes are rounded decimal MB. The ZIPs contain the source or correction request, outline, inspection and check records; the complete presentations are separate PDF downloads.
What if you need to edit the slides?
These results concern the tested international consumer interface. Other Qwen-branded products, accounts or later versions may offer different controls and exports. Qwen Code integrations and third-party presentation services are separate workflows.
Open the file you actually received and check whether text and chart elements can be selected separately. A PDF can preserve the appearance of slides without their editable objects. A PPTX extension alone is also insufficient: each slide can contain a single full-slide image.
- For a fixed presentation: inspect every page and check the software you will present with.
- For changes to wording, figures or charts: verify that those elements are editable. Otherwise, allow time to rebuild the content in a slide editor.
- For a converted file: check the result independently. Conversion does not guarantee that chart data, fonts and layout survive.
We did not convert these PDFs, reconstruct a PPTX or test editing in desktop PowerPoint.
A useful checklist before presenting
- Compare figures with the source. Check them wherever they appear, including headlines and footnotes. A correct outline does not prove the rendered slides kept its numbers.
- Check percentage denominators. For our pilot, 55 of 80 available places means 68.75% utilization. Total attendance is 220 visits; the source does not establish 220 unique people.
- Check chart geometry as well as labels. A bar labelled 40 should reach half an 80-person capacity reference on a zero-based scale. Correct labels on inaccurate bars still mislead.
- Keep the financial scope visible. Our expected $840 is contribution after the two listed costs. Missing marketing and other costs prevent calling it net profit.
- Read the conclusion against the evidence. Week 3 fell from 55 to 50. The later rise to 75 neither proves a cause nor guarantees future attendance.
- Open the downloaded file. Check slide count, legibility, cropped content and the ability to make the edits you need before relying on it.
If you request a correction, list the exact discrepancies and preserve the original. Then review every revised slide and download again; the follow-up may retain errors or introduce new ambiguity.
These checks serve different purposes. Source accuracy, chart accuracy, visual readability and file editability should each pass on their own; a polished cover is not evidence for the others.
Access, limits and common questions
Is Qwen Slides free?
Our September 25 task and September 26 correction completed with the access already available in the signed-in account, without a new purchase or API key. No per-deck charge or remaining-slide allowance was displayed during the task. We did not establish a universal free-slide quota, permanent entitlement or maximum slide count. Check the notices in your own account. Our Qwen pricing guide separates consumer access from developer billing.
Which model creates the slides?
The general chat header on both test dates showed Qwen3.7-Plus. That label does not establish the complete presentation-generation pipeline. We did not independently identify the specialized model rendering each slide.
What if Slides is missing?
Check that you are using the official consumer assistant and inspect its current mode selector after signing in. A similarly named API, local coding workflow or third-party presentation tool can have different controls and exports. Do not assume that changing an API model enables the consumer feature.
Can I use a PDF as source material?
Our presentation test used pasted text, so it does not establish document-to-slide fidelity or attachment limits. For extracting and verifying information from a source PDF first, see our Qwen PDF guide. Recheck the figures after turning them into slides.
How does this relate to Kimi Slides?
Our Kimi Slides guide documents its own workflow and export checks. The Kimi–Qwen comparison contains a separate September 24 nonprofit presentation task with material factual errors in Qwen’s first slides and an unverified Qwen download in that session. The museum task here does not replace that record or establish an overall winner.
Sources and test method
The workflow descriptions come from our September 25, 2026 first test and September 26 assisted-correction session in the official Qwen assistant. Qwen’s consumer overview supplies general product context, but does not establish the download format or editable objects observed in this test.
Test conditions and file inspection — September 25–26, 2026
Both sessions used the signed-in international Qwen consumer assistant. The general header showed Qwen3.7-Plus and Auto on both dates. The account plan and specialized slide-rendering model were not exposed; the header does not identify the whole generation pipeline.
First trial, September 25: we fixed the brief and acceptance checks before submission, sent one fictional source packet, replied only “Confirm” to the first outline and retained the first plan and presentation without regeneration or revision. We inspected all six downloaded pages and the app preview.
Follow-up, September 26: one accepted correction request supplied the correct counts, percentages, totals and financial calculations and requested a financial table. We confirmed the new outline with “Confirm”, kept the first revised output, and inspected all six pages and the PDF structure. Earlier submission attempts were interrupted before a correction appeared in the conversation; the accepted request and confirmation produced one revised deck.
File inspection: each PDF page draws one slide image, with no PDF text objects, fonts or extractable text. Both files use 2688 × 1536-pixel slide images and 720 × 411.43-point PDF pages. The image ratio is exactly 7:4; the rounded PDF page dimensions are approximately 7:4. The files do not contain separate editable text boxes or chart data. No PDF conversion, PPTX reconstruction or desktop PowerPoint editing was tested.
These are two small, related workflow observations, not a performance benchmark or a guarantee for every account. Revising the outline before first generation, other templates and repeated reliability remain untested. The September 30 menu check generated no new deck and does not update these historical results.
See our methodology or report a correction.
For other tasks, return to the Qwen overview, or use the dedicated image-generation, video-generation and Deep Research guides.