Qwen Image 2.1
An open-source image generation model released by Alibaba's Tongyi team. It supports both text-to-image generation and image editing capabilities, and is designed for local deployment scenarios including platforms such as ComfyUI.
Core Facts
Qwen-Image-Edit is an image editing model launched by Alibaba's Tongyi Qianwen team, with capabilities for image generation and local editing.
Qwen Image is a text-to-image model launched by Alibaba.
With the image editing workflow of Qwen Image 2.1, the face-swapping process can be simplified to two images plus a prompt, with no masking operations required.
Timeline (last 90 days)
Searching for Qwen Image 2.1 in Draw Things shows the original teacher model and the Turbo accelerated model, each offered in five quantization precisions; the version without a suffix is the 8-bit version, and the one with Exact is the FP16 version.
Qwen Image 2.1 uses the Qwen Research License, a proprietary research license that opens the weights but restricts direct commercialization.
Qwen Image 2.1 ranks first on the open-source image model leaderboard, surpassing Ideagram 4.
The Draw Things software already supports the Qwen Image 2.1 model.
Less than a month after Qwen Image 2.1 was released, the official team launched an 8-step Turbo version.
The Qwen Research License prohibits selling the model itself as a commercial product or service, bundling and distributing it, or using it as part of commercial infrastructure, and a separate commercial license must be applied for.
Moss Image does not support natural-language semantic editing and cannot directly modify image content through instructions like Qwen Image 2.1; it can only achieve this indirectly through masked inpainting.
A face-swapping LoRA model based on Qwen Image 2.1 appeared in a Reddit community, and its author named it bestfaceswap (abbreviated as bfs).
Qwen Image 2.1 can run on 8GB VRAM in local deployment, but resolution and the number of reference images will be limited, with the number of reference images placing the greatest pressure on VRAM
Qwen Image 2.1 cannot yet replace closed-source image editing products such as GPT or Nano Banana, as it has numerous detail issues and is extremely dependent on prompts.
40 more timeline events
All Facts (20)
Qwen-Image-Edit is an image editing model launched by Alibaba's Tongyi Qianwen team, with capabilities for image generation and local editing.
90%VerifiedQwen Image is a text-to-image model launched by Alibaba.
90%VerifiedWith the image editing workflow of Qwen Image 2.1, the face-swapping process can be simplified to two images plus a prompt, with no masking operations required.
85%VerifiedTesters mentioned that Qwen Image 2.1's image editing and reference image capabilities look promising.
70%VerifiedQwen Image 2.1 is an image generation model with a parameter scale of only 7 billion (7B).
65%VerifiedQwen Image 2.1, an image generation model from Alibaba's Qwen series, has opened early access.
70%UnverifiedUsers who have already installed ComfyUI typically only need to update to a version that supports the model and download the corresponding model weight files to load and use Qwen-Image-2.1 via nodes.
60%UnverifiedCommunity commenters believe that Qwen Image 2.1 only needs to excel in a single dimension (such as editing capability) to have value; otherwise it is just another text-to-image competitor.
60%UnverifiedThe image editing module supports operations such as face swapping while maintaining consistency with the reference image.
60%UnverifiedSearching for Qwen Image 2.1 in Draw Things shows the original teacher model and the Turbo accelerated model, each offered in five quantization precisions; the version without a suffix is the 8-bit version, and the one with Exact is the FP16 version.
50%UnverifiedQwen Image 2.1 uses the Qwen Research License, a proprietary research license that opens the weights but restricts direct commercialization.
50%UnverifiedQwen Image 2.1 ranks first on the open-source image model leaderboard, surpassing Ideagram 4.
50%UnverifiedThe Qwen Research License prohibits selling the model itself as a commercial product or service, bundling and distributing it, or using it as part of commercial infrastructure, and a separate commercial license must be applied for.
50%UnverifiedMoss Image does not support natural-language semantic editing and cannot directly modify image content through instructions like Qwen Image 2.1; it can only achieve this indirectly through masked inpainting.
50%UnverifiedQwen Image 2.1 cannot yet replace closed-source image editing products such as GPT or Nano Banana, as it has numerous detail issues and is extremely dependent on prompts.
50%UnverifiedOfficially released resources include two base models (BF16 and INT8 quantized versions), three text models (BF16, INT8, and FP8), one BF16 VAE, and two PE models supporting image editing and text-to-image generation respectively.
50%UnverifiedIn manga and film/TV storyboard scenarios, Qwen Image 2.1 exhibits detail breakdown issues, while GPT leads significantly in storyboard logic, dialogue attribution, and facial expression rendering
50%UnverifiedUsers with lower-end configurations are advised to use the INT8 quantized model and strictly limit the number of reference images, otherwise VRAM overflow is highly likely
50%UnverifiedThe official recommendation is 40–50 sampling steps; 25 steps can also produce images normally, but 40 or more steps are noticeably better in terms of detail and texture
50%UnverifiedQwen Image 2.1 can run on 8GB VRAM in local deployment, but resolution and the number of reference images will be limited, with the number of reference images placing the greatest pressure on VRAM
50%