The AI arena is free today

Open Superagent
Back to blog
Model Release·Image Generation

Qwen Image 3.0 & 3.0 Pro: Dense Layouts, 10px Text, 12 Languages

Qwen-Image-3.0 targets dense, useful layouts with ~10px text across 12 languages. Pro is $0.04/$0.075 per 1K/2K output; Standard is a flat $0.03. API availability Aug 4, 2026.

Sebastian Crossa
Sebastian Crossa
Co-Founder @ LLM Stats
·2 min read
Qwen Image 3.0 & 3.0 Pro: Dense Layouts, 10px Text, 12 Languages

Alibaba's blog (July 22, 2026) frames the keyword as "Real": rich content, authentic details, deep knowledge. QwenCloud later lists both Pro and the faster Standard tier.

At a Glance

qwen-image-3.0-proqwen-image-3.0
RoleFlagship quality / complex layoutsFaster, lower-cost daily production
Prompt lengthUp to ~4.5k tokens (official claim)Same series claim
Text~10px fine text, 12 languages (self-reported)Same positioning
Max resolution2048x20482048x2048
Max outputs / call66
EditYes (1-3 input images on Pro docs)Yes
Output price (1K)$0.04$0.03
Output price (2K)$0.075$0.03
Input image$0.003$0.003

Pricing from https://docs.qwencloud.com/developer-guides/getting-started/pricing . 1K vs 2K tier is by pixel area (<=2,250,000 = 1K).

What's New

  1. Dense productivity layouts in one pass: newspapers, storyboards, exam papers, multi-cell infographics (Alibaba demoed a 3x3 knowledge grid described with ~3.7k tokens).
  2. Fine text + multilingual fonts (~10px, 12 languages; self-reported).
  3. Pro vs Standard split: Pro for complex / high-fidelity work; Standard for cheaper bulk at the same API shape.

Treat layout and OCR-quality claims as vendor demos until independent image evals land.

Docs describe Standard as a faster variant with the same capabilities as Pro; they do not publish a latency delta. Marketplace rate limits matter for production: Pro lists RPM 5 (concurrency 10); Standard lists RPM 20. Output size on the 3.0 series: custom width/height from 512x512 to 2048x2048, aspect 1:8 to 8:1.

When to Use Which

  • Pro: Newspapers, UI mockups, exam sheets, LaTeX-heavy figures, multilingual posters where text accuracy matters.
  • Standard: High-volume posters / web / UI drafts where $0.03 flat and speed matter more than the last bit of fidelity.

Sources

Questions

Frequently Asked Questions

  • Alibaba Cloud blog July 22, 2026. QwenCloud changelog lists both qwen-image-3.0-pro and qwen-image-3.0 as new on August 4, 2026 (use Aug 4 for API availability).
  • Pro: $0.04 (1K out) / $0.075 (2K out). Standard: $0.03 per image at 1K or 2K. Input images $0.003 each.

  • Not required for this short launch note. Capability claims above are self-reported from Alibaba / QwenCloud.

Continue Reading