Qwen-Image-3.0-Pro is the flagship tier of Alibaba's Qwen-Image-3.0 image generation and editing model. A single model handles both text-to-image generation and image-to-image editing from one to three reference images, returning up to six outputs per request at resolutions up to 2048x2048. It uses a roughly 20-billion-parameter MMDiT (multimodal diffusion transformer) architecture and supports prompts of about 4.5K tokens, legible text down to around 10 pixels, and native rendering of 12 languages with a wide range of fonts. Through Qwen Cloud it costs about $0.04 per image, versus roughly $0.03 for the standard tier, positioning it for commercial work such as advertising creative and multilingual poster production.
Parameters
20B
Context Window
4.5K
License
Proprietary
Release Date
2026-07-21
Benchmark Performance
AA Intelligence Index
—
LMArena Elo
—
HLE
—
ARC-AGI-2
—
SWE-bench Verified
—
GPQA Diamond
—
MMLU-Pro
—
LiveCodeBench
—
AIME 2025
—
MATH-500
—
API Pricing
API pricing for this model is not yet available
Strengths
- •High-fidelity text rendering that stays legible down to around 10 pixels
- •Native rendering of 12 languages and a wide range of fonts
- •Generation and editing handled by a single model through one API
Weaknesses
- •Maximum output resolution capped at 2048x2048
- •Closed model with no published weights
- •Higher per-image cost than the standard 3.0 tier
Use Cases
- •Advertising creative with heavy typographic elements
- •Batch generation of multilingual posters and e-commerce banners
- •Reference-image editing of product photography