whatmodel.sh / image / Qwen-Image-Edit-2511

Qwen-Image-Edit-2511

+ contribute
Update model informationAdd a place to run itAdd a test resultPropose a benchmarkCorrect a benchmark scoreEdit data on GitHub

Qwen-Image-Edit-2511 is an open-weights image model from Alibaba, released December 2025. Its published size is 20.43B parameters. Licence: Apache 2.0.

Model information sources

Release date ↗ · Parameter count ↗

MMDiT transformer only: HF Hub safetensors total for the official repo (pinned revision), 20,430,401,088 = the sum of the transformer/ shard headers. The Qwen2.5-VL text encoder and the VAE are not included (convention of qwen-image = 20). The card says the model is 'Built upon our 20B Qwen-Image model'. Observed 2026-10-07.

57.7±1.8#57 of 144 image modelsfrom 8 benchmark results across 8 sources

Score within the LMArena text-to-image comparison set only.

$ how it compares

The image models ranked either side of Qwen-Image-Edit-2511, out of 197 with a score, then more from Alibaba.

#modelscorewhereRAMreleased
64Wan 2.6 Image Alibaba58±4api $0.030/image2025-12
65HunyuanImage 2.1 Tencent58±2api $0.10/image +local~64GB2025-09
66Qwen-Image-Edit-2511 Alibaba58±2api $0.030/image +local~37GB2025-12
67Recraft V4.1 Flash Recraft57±3api $0.0070/image2026-09
68HunyuanImage 3.0 Tencent57±2api $0.080/image +local~49GB2025-09
more from Alibaba
11Qwen-Image-3.0-Pro Alibaba75±2api $0.040/image2026-07
18Qwen-Image-2.1 Alibaba69±2local~12GB2026-09
20Qwen-Image-2.0-Pro (2026-06-22) Alibaba67±2api $0.075/image2026-06

see all image models · find the ones that fit your machine

$ what it's good at

image to image / editing▁▁·▃▂▃█▃▃▄▁▁ #33 of 62 · 67% of its score · 4 tests

Beats FLUX.2 [flex], Imagen 4, GPT Image 1 Mini here, though they rank higher overall.

testresultvs the others in its rundifficultycountswhy
Edit-Compass (EN, overall)53.8% ✎▃▃▂·▃▆▂█·▃·▃ beats 16/20hard · avg 45%▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇ 36%at its level
Arena image edit1234.0 ✎▂▁▁▂▆▇█▄▆▅▁▂ beats 18/53avg 1153▇▇▇▇▇▇▇ 18%typical
GEditBench v2 (overall Elo)1038.0 ✎▃▂····▂█▂▃▂▃ beats 10/14avg 956▇▇▇ 7.1%blunt
MultiTextEdit cross-lingual text-in-image editing (SemAvg, LVM judge)64.3% ✎▄·▄▄▄▄··█▄·▄ beats 6/8medium · avg 50%▇▇ 5.3%at its level
other tests4 not in a task · 32% of its score
testresultvs the others in its rundifficultycountswhy
ChartSync chart editing (Overall)60.3% ✎▂··▂█▂▄▂▂·▂▂ beats 7/12medium · avg 41%▇▇▇▇▇▇▇ 16%at its level
InfoEdit infographic editing (average success rate)0.4% ✎▄··▄█····▄▄▄ beats 0/6very hard · avg 1.4%▇▇▇ 8.5%sharp · too hard for it
PaintBench precise editing (mIoU, 20-task average)6.7% ✎▃·····█▅▃·▃▅ beats 6/9very hard · avg 4.3%▇▇ 4.0%too hard for it
ViGoR-Bench zero-shot visual reasoning, image models (Result Avg)25.8% ✎▂···▂·▂█▆▅·▂ beats 5/15hard · avg 31%▇ 2.9%at its level

$ where to run it

Through an API

api · fal.ai · $0.030/image · fal.ai ↗ · price sourceUSD for one 1-megapixel output: fal bills this endpoint $0.03 per megapixel (price = rate × 1 MP; fal's rounding for 1024×1024 ≈ 1.05 MP is not stated; larger outputs cost proportionally more). Product: fal.ai serverless image editing. API model: fal-ai/qwen-image-edit-2511 (fal describes it as the endpoint for Qwen's Image Editing 2511 model). Settings: one input image, 1 MP output, one image. Price basis: published per-megapixel list rate × 1 MP. Alibaba Model Studio's own qwen-image-edit/-plus API models are not used because Alibaba does not state which open checkpoint they correspond to. Currency: USD. Not an Alibaba direct tariff. fal.ai official model page, observed 2026-09-26.

Other ways to run it

  1. Run it as published · needs ~36.6 GB · from github.com/leejet
    sd-cli --diffusion-model models/qwen-image-edit-2511-Q4_K_M.gguf --vae models/qwen_image_vae.safetensors --llm models/qwen_2.5_vl_7b.safetensors --cfg-scale 2.5 --sampling-method euler -v --offload-to-cpu --diffusion-fa --flow-shift 3 -r input.png -p "change 'flux.cpp' to 'edit.cpp'" --model-args qwen_image_zero_cond_t=true

Official weights and documentation: huggingface.co/Qwen ↗

$ how the score was worked out · Edit-Compass & EditReward-Com… 36%, LMArena leaderboard dataset 18%, ChartSync: A Benchmark for Vi… 16% and 5 more

Each result implies a score on its own, after adjusting for how hard its test is and who ran it. The score is their weighted average, with a small pull toward the middle for thinly tested models. Each test's weight is its counts column above; here it's added up by who ran the tests. How it works

sourceresultsshare of the scorelicence
Edit-Compass & EditReward-Compass: A Unified Benchmark for Image Editing and Reward Modeling1▇▇▇▇▇▇▇▇▇▇▇ 36%CC-BY-4.0
LMArena leaderboard dataset1▇▇▇▇▇ 18%CC BY 4.0
ChartSync: A Benchmark for Visuo-Logical Cascading Chart Editing (arXiv:2607.10301v1)1▇▇▇▇▇ 16%CC-BY-4.0
InfoEdit: Probing Global Layout Reasoning in Infographic Editing (arXiv:2609.33286v1)1▇▇▇ 9%CC-BY-4.0
GEditBench v2: A Human-Aligned Benchmark for General Image Editing1▇▇ 7%CC-BY-4.0
MultiTextEdit: Benchmarking Cross-Lingual Degradation in Text-in-Image Editing (arXiv:2605.08163v2), Table 31▇▇ 5%CC BY 4.0
PaintBench: Deterministic Evaluation of Precise Visual Editing (arXiv:2606.00188v1)1▇ 4%CC-BY-4.0
ViGoR-Bench: How Far Are Visual Generative Models From Zero-Shot Visual Reasoners? (arXiv:2603.25823v1)1▇ 3%CC-BY-4.0
prior · pull toward the middle▇ 1%
weighted average= 57.7

$ improve this page

Update model informationAdd a place to run itAdd a test resultPropose a benchmarkCorrect a benchmark scoreEdit data on GitHub

Use a short GitHub form or edit the data through a pull request. Every number needs a source; here's which ones we can use.

data updated · sources and licences · json