Qwen-Image: Crafting with Native Text Rendering
We are excited to introduce QwenImage — a **20B MMDiT foundation model** for image generation and editing that brings **breakthrough advancements in complex text rendering** and **precision-driven image editing**.
Qwen-Image: Crafting with Native Text Rendering
We are excited to introduce Qwen-Image — a 20B MMDiT foundation model for image generation and editing that brings breakthrough advancements in complex text rendering and precision-driven image editing.
You can try the latest model now via Qwen Chat by selecting "Image Generation".
Key Features
1. Superior Text Rendering
Qwen-Image delivers unmatched accuracy when rendering complex text, handling:
- Multi-line layouts
- Paragraph-level semantics
- Fine-grained typographic details
It supports both alphabetic languages (e.g., English) and logographic languages (e.g., Chinese) with exceptional clarity and fidelity.
2. Consistent Image Editing
Powered by an enhanced multi-task training paradigm, Qwen-Image maintains both semantic meaning and visual realism throughout complex editing workflows.
3. Cross-Benchmark Leadership
Tested across multiple public benchmarks, Qwen-Image consistently outperforms competing models in both image generation and image editing tasks — solidifying its position as a true foundation model for creative AI.
Performance Highlights
We conducted extensive evaluations of Qwen-Image on:
- Image Generation Benchmarks: GenEval, DPG, OneIG-Bench
- Image Editing Benchmarks: GEdit, ImgEdit, GSO
- Text Rendering Benchmarks: LongText-Bench, ChineseWord, TextCraft
The results speak for themselves:
- State-of-the-art performance in both generation and editing tasks
- Significant lead in Chinese text rendering and other complex multilingual scenarios
- Outstanding versatility across photorealistic, artistic, anime, and minimalist styles
Creative Versatility
Qwen-Image adapts seamlessly to a variety of artistic and commercial needs:
- From real-world photography to impressionistic painting
- From anime-inspired illustrations to clean, minimal graphic design
For editing, it supports:
- Style transfer
- Object addition or removal
- Detail enhancement
- Text modification within images
- Character pose adjustments
This makes professional-grade editing accessible to designers, marketers, and everyday creators.
A Step Towards Open, Sustainable AI
Our goal with Qwen-Image is to:
- Lower the technical barrier to high-quality visual content creation
- Empower developers, artists, and storytellers with next-generation creative tools
- Foster a transparent, community-driven ecosystem for generative AI
We welcome the active participation and feedback of the community to help shape Qwen-Image into an even more capable and inclusive creative platform.
👉 Try Qwen-Image now: https://www.qwenimagen.com/