Qwen-Image-2.1
AlibabaOpenqwen-research
Qwen-Image-2.1 is an AI image model from Alibaba. On LMArena’s Text-to-image leaderboard it ranks 19 of 82 with 1223±7 from 11,441 votes. On LMArena’s Image editing leaderboard it ranks 19 of 60 with 1362±6 from 36,713 votes. License: qwen-research.
Key figures
- Developer
- Alibaba
- License
- qwen-research
- Released (per AA)
- Sep 2026
Who it is for (from the data)
- Solid quality with room to choose: #19 of 82 on Text-to-image, in the top half.
- Generating and editing images: it is ranked on both boards.
- Self-hosting or on-premise use: LMArena lists its license as qwen-research.
- Existing Alibaba users: this vendor has 11 models on Text-to-image.
Neighbours on Text-to-image
| # | Model | Score | Votes | AA Elo | License | Δ |
|---|---|---|---|---|---|---|
| 16 |  Gemini 3 Pro Image 2K (Nano Banana Pro) Google | 1248 | 175,734 | – | Proprietary | +25 |
| 17 |  GPT Image 1.5 High Fidelity OpenAI | 1237 | 169,927 | – | Proprietary | +14 |
| 18 | Gemini 3 Pro Image Preview (Nano Banana Pro) Google | 1234 | 86,464 | – | Proprietary | +11 |
| 19 | Qwen Image 2.1 Alibaba | 1223 | 11,441 | 1042 | qwen-research | |
| 20 |  Ideogram 4.0 Quality Ideogram | 1205 | 60,591 | – | Ideogram Open Model | -18 |
| 21 | Qwen Image 2.0 Pro 2026-06-22 Alibaba | 1192 | 13,096 | – | Proprietary | -31 |
| 22 |  UNI 1.1 Max Luma AI | 1189 | 14,072 | – | Proprietary | -34 |
Alternatives · More from Alibaba
| # | Model | Score | Votes | AA Elo | License |
|---|---|---|---|---|---|
| 13 | Qwen Image 3.0 Pro Alibaba | 1255 | 11,440 | 1095 | Proprietary |
| 21 | Qwen Image 2.0 Pro 2026-06-22 Alibaba | 1192 | 13,096 | – | Proprietary |
| 41 | Wan 2.6 T2I Alibaba | 1137 | 233,292 | – | Proprietary |
| 44 | Qwen Image 2512 Alibaba | 1125 | 98,288 | – | Apache 2.0 |
| 48 | Wan 2.5 T2I Preview Alibaba | 1116 | 279,229 | – | Proprietary |
Alternatives · Top open-weight models
| # | Model | Score | Votes | AA Elo | License |
|---|---|---|---|---|---|
| 20 | Ideogram 4.0 Quality Ideogram | 1205 | 60,591 | – | Ideogram Open Model |
| 27 |  Cosmos3-Super-Text2Image (Agentic) NVIDIA | 1165 | 10,805 | – | OpenMDW-1.1 |
| 33 | Cosmos3-Super-Text2Image NVIDIA | 1152 | 18,485 | – | OpenMDW-1.1 |
| 35 |  Hunyuan Image 3.0 Tencent | 1150 | 180,518 | 989 | tencent-hunyuan-community |
| 38 |  FLUX 2 Dev Black Forest Labs | 1146 | 77,988 | – | flux-non-commercial-license |
How to read the scores
- LMArena score: people blind-pick the better of two anonymous images for the same prompt, and the votes are fitted into an Elo-style rating; "±" is the 95% confidence interval.
- Text-to-image and image edit are separate boards, so scores are not comparable across them; AA Elo comes from Artificial Analysis on a different scale.
- If the gap between two models is smaller than their combined intervals, the order is not statistically reliable; treat them as the same tier. More votes mean narrower intervals.
FAQ
Where does Qwen-Image-2.1 rank among AI image models?
Text-to-image rank 19 (1223); Image editing rank 19 (1362) (LMArena, data as of Oct 26, 2026).
Is Qwen-Image-2.1 open-weight?
LMArena lists its license as qwen-research (not proprietary); check the license text for commercial-use terms.
Can Qwen-Image-2.1 edit images?
Yes. Qwen-Image-2.1 is #19 of 60 on the LMArena image-edit board (1362, 36,713 votes).
Who makes Qwen-Image-2.1?
According to LMArena, Qwen-Image-2.1 is made by Alibaba.
How reliable is Qwen-Image-2.1's score?
The scores are based on 11,441 votes on Text-to-image (95% CI ±7); 36,713 votes on Image editing (95% CI ±6). More votes and narrower intervals mean a more stable rank.
How far is Qwen-Image-2.1 from #1 on Text-to-image?
#1 on Text-to-image is GPT Image 2.5 Sunburst (1425); Qwen-Image-2.1 scores 1223, 202 points behind.
Sources
LMArena Text-to-image data published on Oct 7, 2026.
LMArena Image editing data published on Oct 20, 2026.
17NAS syncs the upstream leaderboards daily; this page's data was synced on Oct 26, 2026.
Price, resolution and speed have no reliable public data yet and are omitted; all values come from the snapshot, blanks show –.