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)
Qwen Image 2.1 could correctly follow color and spatial position instructions (such as a red cube on the left and a blue sphere on the right on a table).
In most generated images, Qwen Image 2.1 repeatedly produced purple mark artifacts resembling printing marks.
In the test, Qwen Image 2.1 could correctly spell basic words but could not accurately generate the number of objects requested in the prompt (such as 5 apples).
Bilibili creator Luke's DevLab used a single RTX 5090 to run a stress test of 32 prompt sets on Qwen Image 2.1 locally.
The creator confirmed that Qwen Image 2.1 can generate uncensored content.
In a scene of a cup being filled with water, Qwen Image 2.1 got the water amount clearly wrong, with too much water ruining the composition.
Qwen Image 2.1 performed well on perspective and repeated structures (corridor arches, grid floors), with elements aligned neatly.
Qwen Image 2.1 performed poorly in tests of fine geometric structures such as bicycle wheel spokes, with lines showing spiral branching and chaos.
Qwen Image 2.1 performed poorly on realistic human anatomy (specified poses, hand details) but was noticeably better on anime-style human figures.
Qwen Image 2.1 uses the Qwen Research License, a proprietary research license that opens the weights but restricts direct commercialization.
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%UnverifiedQwen Image 2.1 performed poorly on realistic human anatomy (specified poses, hand details) but was noticeably better on anime-style human figures.
50%UnverifiedQwen Image 2.1 performed poorly in tests of fine geometric structures such as bicycle wheel spokes, with lines showing spiral branching and chaos.
50%UnverifiedIn the test, Qwen Image 2.1 could correctly spell basic words but could not accurately generate the number of objects requested in the prompt (such as 5 apples).
50%UnverifiedIn a scene of a cup being filled with water, Qwen Image 2.1 got the water amount clearly wrong, with too much water ruining the composition.
50%UnverifiedQwen Image 2.1 could correctly follow color and spatial position instructions (such as a red cube on the left and a blue sphere on the right on a table).
50%UnverifiedIn most generated images, Qwen Image 2.1 repeatedly produced purple mark artifacts resembling printing marks.
50%UnverifiedQwen Image 2.1 performed well on perspective and repeated structures (corridor arches, grid floors), with elements aligned neatly.
50%UnverifiedThe creator confirmed that Qwen Image 2.1 can generate uncensored content.
50%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%