Executive Summary
Qwen-Image-2.0 was announced on 10 February 2026 by Alibaba's Qwen team, claiming superior text rendering and native 2K output. The model is currently in invitation-only beta; the publicly available qwen-image-max on DashScope was tested instead. This report compares it head-to-head with Gemini nano-banana-pro-preview (the model behind the /gemini-images skill) across four use cases matching the existing image generation templates.
Gemini wins 3–1 overall. Text accuracy, style adherence, and resolution make it the clear choice for the current workflow.
Comic Strip
Gemini
Infographic
Gemini
Presentation
Qwen
Text Rendering
Gemini
1 Comic Strip
A 4-panel comic about a developer over-engineering a simple button with abstraction layers. Hand-drawn colored-pencil style in brand colours.
Qwen qwen-image-max
928×1664 · 2,515 KB
Gemini nano-banana-pro-preview
~1500×2000 (2K) · 1,613 KB
Analysis
| Criteria | Qwen | Gemini |
|---|---|---|
| Style match | Semi-realistic digital illustration. Not the hand-drawn pencil aesthetic requested. | Nails the whimsical colored-pencil look. Exactly on-brand. |
| Storytelling | 3 panels, adequate but compressed narrative. | 5 panels with better arc. Added superhero cape gag — creative embellishment. |
| Speech bubble text | Legible, fewer words. | More text, all legible, funnier dialogue. |
| Code/class name labels | Garbled and blurry on the Jenga blocks. | Readable: "AbstractFactoryProviderManagerService", etc. |
| Brand colours | Generic blue/teal. Not distinctly on-brand. | Closer to the turquoise/orange palette specified. |
| Copyright footer | Present and correct. | Present and correct. |
Winner: Gemini — better style adherence, storytelling, and text accuracy on labels.
2 Infographic
"5 Signs Your AI Strategy Is Actually Working" — portrait infographic with puzzle-piece rocket centrepiece, 5 numbered points, and newsletter CTA footer.
Qwen qwen-image-max
928×1664 · 2,281 KB
Gemini nano-banana-pro-preview
~1856×2304 (2K) · 1,554 KB
Analysis
| Criteria | Qwen | Gemini |
|---|---|---|
| Title | Present and bold. | Present, bolder, with better styling. |
| Central motif | Puzzle rocket present, reasonable quality. | Excellent puzzle rocket, more whimsical and engaging. |
| Text accuracy | Multiple typos: "ausks" (asks), "juss" (just), "proj oject" (project), "dalloring" (delivering). Dealbreaker. | All text correct. Zero typos across all 5 points. |
| Layout | Text down one side only. Less dynamic composition. | Text arranged around centrepiece with supporting illustrations (people, gears, charts, rubbish bin). |
| Footer | Says "2024" — wrong year. | Correct: "2026", includes chrismdp.com. |
Winner: Gemini — decisive. Qwen's text typos and wrong year in the footer are disqualifying for professional use.
3 Presentation Slide
Widescreen 16:9 slide showing an S-curve "AI Adoption Curve" with three labelled phases and an orange "You are here" arrow. Minimalist, max 10 words.
Qwen qwen-image-max
1664×928 · 1,983 KB
Gemini nano-banana-pro-preview
1920×1080 (1K) · 199 KB
Analysis
| Criteria | Qwen | Gemini |
|---|---|---|
| Data visualisation | Excellent. Proper S-curve with dot clusters showing density changes. Reads like a real chart. | Abstract swooping line. More artistic but less informative as a data viz. |
| Text labels | All three labels perfect and well-placed on the x-axis. | All three labels present and readable. |
| "You are here" arrow | Orange arrow pointing down at the correct position on the curve. | Orange arrow present but positioned at the bottom, disconnected from the curve. |
| Wobbly border | Turquoise sketch border, present. | Turquoise sketch border, present. |
| Minimalism | Clean, tight, information-dense without clutter. | Spacious, visually appealing but curve itself is less useful. |
Winner: Qwen — the S-curve with clustered dots is a genuinely better data visualisation. More useful as an actual presentation slide.
4 Text Rendering (Quote Card)
Peter Drucker quote card with precise typography, turquoise border, orange flourishes, and "chrismdp.com" below. This tests pure text rendering — Qwen's claimed strength.
Qwen qwen-image-max
1328×1328 · 1,621 KB
Gemini nano-banana-pro-preview
1024×1024 (1K) · 106 KB
Analysis
| Criteria | Qwen | Gemini |
|---|---|---|
| Quote text | Perfect rendering, elegant serif font. | Perfect rendering, elegant serif font. |
| Attribution | "— Peter Drucker" — correct. | "— Peter Drucker" — correct. |
| URL text | "chrasmedp.com" — misspelled. Letters transposed. | "chrismdp.com" — correct. |
| Orange flourishes | Present, small and subtle. | Present, more ornate and decorative. |
| Border | Turquoise border present. | Turquoise border present. |
| Overall composition | Dark surround outside the border feels off-brand. More compact. | Clean, spacious, white surround. On-brand. |
Winner: Gemini — ironic given Qwen's text-rendering claims. The misspelled URL ("chrasmedp.com") is exactly the kind of error that can't ship.
Conclusions & Recommendations
Where Gemini excels
- Text accuracy — Zero errors across all 4 tests. Qwen had typos in 2 of 4 (infographic and quote card), which is a dealbreaker for professional content.
- Style adherence — Consistently delivers the hand-drawn colored-pencil aesthetic that the brand templates specify. Qwen defaults to a more photorealistic digital illustration style.
- Resolution — Native 2K output (up to 2304px tall for infographics). Qwen caps at 1664×928, which can't meet the infographic template spec of 1856×2304.
- Creative interpretation — Added the superhero cape in the comic, the rubbish bin for "killed projects" in the infographic. More inventive with the brief.
Where Qwen excels
- Data visualisation — The presentation slide S-curve was genuinely better composed, with dot clusters that communicate density changes intuitively.
- File weight — Consistently higher-resolution output files (1.6–2.5 MB vs Gemini's 0.1–1.6 MB), suggesting more detail per pixel.
- Price — $0.075/image vs Gemini's usage-based pricing. 100 free images on signup.
Key limitations of Qwen-Image-Max
- Max resolution 1664×928 — Can't produce the 1856×2304 infographics or 1920×1080 presentation slides at native resolution.
- Text rendering errors — Despite marketing claims about superior text handling, produced "ausks", "juss", "dalloring", and "chrasmedp.com" in testing.
- Style rigidity — Tends toward photorealistic rather than adapting to the requested hand-drawn aesthetic.
- No inline image return — Returns a temporary URL (expires in 24h) rather than base64 data, adding complexity to automation.
Recommendation
Stick with Gemini for the /gemini-images skill. Qwen-Image-Max is not a viable replacement for any of the current template types. The text accuracy problems alone are disqualifying — you can't publish infographics with "dalloring" instead of "delivering".
Worth revisiting when Qwen-Image-2.0 gets a public API (currently invitation-only on Bailian). The 2.0 model claims 2048×2048 native output and improved text rendering, which could address the two biggest gaps. Monitor the DashScope model list for a qwen-image-2.0 or updated qwen-image-max model name.
One possible niche use: If you ever need chart/diagram-style slides where data visualisation matters more than brand aesthetics, Qwen could serve as a secondary option.