Qwen-Image
Alibaba's open-source 20B MMDiT image foundation model under Apache 2.0, best-in-class for multilingual text rendering including Chinese.
Overview
Qwen-Image is the first image generation foundation model from Alibaba's Qwen team, released open-source under Apache 2.0 in August 2025 (later refreshed with the Qwen-Image-2512 weights). The 20B-parameter MMDiT (Multimodal Diffusion Transformer) is known for commercial-grade complex text rendering in both English and Chinese and precise image editing, and has ranked among the top open-source image models on public arenas. Weights are freely available on Hugging Face and ModelScope, with a broad LoRA and fine-tune ecosystem. It is distinct from Alibaba's Qwen chat models and the Tongyi Wanxiang generator.
Pricing
Pricing shown for reference only. These figures reflect RECATOOLS research as of 24 Jul 2026 and may be out of date or incomplete. This is not financial or purchasing advice — always confirm the current price on the provider’s official website before making any decision.
ASEAN Perspective
Qwen-Image in Southeast Asia
ASEAN-region availability and pricing notes coming soon. Drop the editorial team a note via /contact/ if you can supply local context (Singapore/Malaysia/Indonesia/Thailand/Vietnam).
What this is for: Open-source text-to-image generation and editing with standout multilingual text rendering, strong for posters and Chinese/English typography.
Who this is for: Developers, researchers, and studios wanting a permissively licensed image model they can self-host and fine-tune.
Availability: Free open weights under Apache 2.0 on Hugging Face and ModelScope; also available via hosted APIs. Self-hostable.
What people say
The release of Qwen-Image was a significant event for open-weight models. The Apache-2.0 20B MMDiT was quickly given native ComfyUI support and picked up a broad LoRA/quantized-GGUF ecosystem on Civitai and Hugging Face, and Alibaba reports the refreshed Qwen-Image-2512 ranks as the strongest open-source entry across 10,000+ blind AI Arena rounds while staying competitive with closed systems. Practitioner consensus on Reddit, ComfyUI, and diffusion blogs is that its standout strength is text rendering. It handles long, multi-line paragraphs in English and Chinese—typography that previous open models mangled—and its general quality is in the neighborhood of Flux Kontext Pro, Imagen 3, and Ideogram 3.
The caveats are the familiar open-model ones. At 20B it is heavy: full-precision runs demand serious VRAM, so most people rely on fp8/GGUF quantizations that trade some fidelity for feasibility. Community feedback also notes it can lag dedicated aesthetic-tuned checkpoints on certain artistic styles and human anatomy, and that arena wins don't always translate to every prompt. It's easy to confuse with Alibaba's separate Qwen chat models and the Wan/Tongyi Wanxiang generators, and the strongest results often require the right community LoRAs and settings rather than working perfectly out of the box.
Summary of public user & expert reviews, compiled by RECATOOLS.
About this listing
This entry was compiled from publicly available data including Qwen-Image's official website, press releases, documentation, and reputable third-party publications. RECATOOLS is not affiliated with Qwen-Image unless explicitly stated.
Third-party AI tools update their pricing, features, availability, and policies frequently. Information here may be outdated by the time you read this — we make reasonable efforts to keep listings current, but cannot guarantee absolute accuracy.
For the latest details, please refer to Qwen-Image directly →
Spotted something out of date? Suggest an update →
More in Image Generation
