🚨 Alibaba released Qwen-Image-3.0 today I attached a few examples, images 2nd through 4th are generated by it The text rendering and layout on that academic paper image is very very impressive, and it seems to have improved world knowledge
Qwen Image 3.0
Qwen Image 3.0 係阿里巴巴第三代圖像模型,專注令生成圖像真正可以用喺實際工作上。支持最長 4500 token 嘅超長指令,可以一次生成資訊密集型內容,仲有原生 12 種語言同 20+ 字體嘅文字渲染能力——喺報紙版式、多欄資訊圖、含數學公式嘅學術論文、UI 稿同電商圖像上表現出色。
🎯 探索50+模型Qwen Image 3.0 文生圖畫廊
探索 Qwen Image 3.0 嘅 AI 圖像創作,阿里巴巴第三代圖像模型。專為資訊密集型輸出打造,支持最長 4500 token 超長指令同原生 12 種語言文字渲染——由精細資訊圖到照片級真實場景。

舞蹈人像
一位齊腰長臟辮嘅黑人女性旋轉中,臟辮隨離心運動散開,纏有金線、寶貝殼同藍色珠子,戲劇性頂光營造光環效果。
“A Black woman with waist-length loc'd hair caught mid-spin in a dance, her locs fanning out radially following centrifugal motion, some locs wrapped with gold thread, some with cowrie shells, some with blue beads. Individual locs are clearly separated and countable. She is wearing a flowing white dress also caught in the spin motion. Droplets of water flying off the loc tips against a dark background, lit by a single overhead spotlight creating a halo effect”
用 Qwen Image 3.0 進行 AI 圖像編輯
用 Qwen Image 3.0 轉換、重塑並重新構想你嘅圖像。由照片級風格遷移到多圖合成,以及支持 12 種語言嘅圖內文字翻譯。


海灘風格遷移
Qwen Image 3.0 YouTube 影片
睇吓展示 Qwen 圖像模型能力嘅社區教程同評測
- 🚀 Introducing Qwen-Image-2.0 — our next-gen image generation model! - Qwen
- Qwen-Image 2.0: Professional Infographics + Photorealism in ONE AI Model - Fahd Mirza
- Qwen Image 2.0 — Qwen Next-gen Image Generation Model - Codedigipt
- "Qwen Image 2.0 Review: Insane Image & Text Rendering + Editing Beast!" - kintu
- Qwen-Image 2.0 is HERE! - Sudo AI
Qwen Image 3.0 YouTube 影片
睇吓展示 Qwen 圖像模型能力嘅社區教程同評測
Qwen Image 3.0 喺 X 上嘅熱門評價
睇吓大家喺 X(推特)上點樣評價 Qwen Image 3.0
🚀 Introducing Qwen-Image-2.0 — our next-gen image generation model! 🎨 Your imagination, unleashed. ✨ Type a paragraph → get a pro slides ✨ Describe a scene → get photoreal 2K magic ✨ Add text → it just works (no more glitchy letters!) ✨ Key upgrades: ✅ Professional Show more
A quick update — we’ve fixed a Qwen-Image 2.0 bug in Qwen Chat that impacted: • Classical Chinese poem ordering in image generation • Character consistency during image editing ✅Patch is live now! chat.qwen.ai/?inputFeature=… Go test it out and drop us your feedback.
Qwen-Image 2.0 was released by @Alibaba_Qwen, marking a significant advancement in visual AI. Its most notable upgrade is the ability to render precise, legible text within images, including English and Chinese, enabling the creation of professional infographics, posters, and Show more
qwen has just announced a new image model, qwen-image-2.0, which merges qwen-image and qwen-edit into a single base. It ranked third in the generate ability test and second in the editing ability test. Not yet open source; currently available with API and chat webpage.
Qwen-Image is the new open weights image champion model, surpassing HiDream-I1-Dev with quality rivaling Imagen 3 and FLUX.1 Kontext [pro] in the Artificial Analysis Image Arena! Following the recent release of Wan2.2 A14B (the leading open weights video model), Alibaba has Show more
Qwen Image Edit 2509 is the new leading open weights image editing model, ranking #3 overall in the Artificial Analysis Image Editing Arena and introducing multi-image editing capabilities! The latest release from Alibaba Qwen trails only Gemini 2.5 Flash (Nano-Banana) and Show more
🇨🇳Alibaba’s Qwen-Image-2.0 Launches as Next-Gen Image Generation Model Alibaba’s Qianwen team has launched the next-generation image-generation foundation model Qwen-Image-2.0. The model’s API is now available for invited testing on Alibaba Cloud Bailian, and developers can also Show more
🚀 Qwen Image just got a big update on WaveSpeedAI! @Alibaba_Qwen ⚡ Generate ultra-accurate images in as fast as 5 seconds — just $0.02 per image and API ready! 🧠 Say goodbye to gibberish — Qwen Image nails text rendering like no other. Try it now 👉 wavespeed.ai/models/wavespe… Show more
Qwen-Image-2.0、驚異のフォトリアリズム能力で注目。 GTA画像生成例も公開され、プロンプト忠実度、編集、写実性で既存モデル凌駕の高性能。 開発チームは深層理解と高忠実度生成を単一7Bアーキテクチャで統合と明言。 現在公式サイトで利用可能、オープンソース化に期待集まる。 #QwenImage2 Show more
什麼是 Qwen Image 3.0
阿里巴巴第三代圖像模型,專為資訊密集、即用型輸出打造
- · 014.5K最大Token
- · 0212語言
- · 0320+字體
- · 04第3代最新一代
Qwen Image 3.0 係阿里巴巴 Qwen 圖像模型嘅第三代,專注令生成圖像夠實用、可以直接用嚟做嘢。佢支持最長 4500 token 嘅超長指令——大約係上一代嘅 4.5 倍——可以一次生成資訊密集型內容,仲原生渲染 12 種語言同 20 幾種字體嘅文字,喺報紙、資訊圖、學術論文等密集版式上表現出色。
Qwen Image 3.0 功能特性
了解 Qwen Image 3.0 為資訊密集、專業級輸出而生嘅能力
- Feature 01 / 08
4.5K 超長指令
支持最長 4500 token 嘅指令——大約係上一代嘅 4.5 倍——可以一次生成複雜、資訊密集嘅圖像。
- Feature 02 / 08
12 種語言原生文字
原生渲染 12 種語言嘅文字,令全球化多語言版式嘅排版準確清晰。
- Feature 03 / 08
20+ 字體渲染
還原 20 幾種字體嘅忠實字形,可生成商業級海報、標題同正文。
- Feature 04 / 08
報紙同文檔版式
處理報紙版面、多欄文檔等密集、文字繁多嘅構圖,版式結構清晰、易睇。
- Feature 05 / 08
專業資訊圖
一句提示詞就可以生成數據豐富嘅資訊圖網格同演示視覺,標籤準確、細字清晰可讀。
- Feature 06 / 08
數學同技術符號
繪製公式、幾何圖形同技術圖示,適用於學術論文同教育材料。
- Feature 07 / 08
世界知識同介面還原
憑住內置世界知識,還原網頁、直播畫面等真實世界介面。
- Feature 08 / 08
生成同編輯
喺單一工作流入面由文字生成再用圖像編輯精修,由初稿到可以直接用嘅成品一氣呵成。
如何使用 Qwen Image 3.0 文生圖
三步生成資訊密集、多語言文字精準嘅圖像
輸入提示詞
詳細描述你想要創建嘅圖像。Qwen Image 3.0 支持最長約 4500 token 嘅超長指令,可以包含詳細嘅佈局指令、文字內容同風格偏好。