whatmodel.sh / image / Show-o2 7B

Show-o2 7B

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

Show-o2 7B is an open-weights image model from Show Lab, NUS, released June 2025. Its published size is 7B parameters. Licence: apache-2.0.

Model information sources

Release date ↗ · Weights licence ↗ · Parameter count ↗

7B = the language-model backbone (Qwen2.5-7B-Instruct) as the makers state it ('7B LLM parameters'); the flow head, spatial(-temporal) fusion/projector layers and the external Wan2.1 3D causal VAE are not included, and no total is published.

41.5±2.5#103 of 144 image modelsfrom 4 benchmark results across 3 sources

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

$ how it compares

The image models ranked either side of Show-o2 7B, out of 197 with a score, then more from Show Lab, NUS.

#modelscorewhereRAMreleased
136OmniGen2 BAAI42±2api $0.15/image +local~17GB2025-06
137Stable Diffusion 3.5 Large Stability AI42±1api $0.065/image +local~5.4GB2024-10
138Show-o2 7B Show Lab, NUS42±3local~23GB2025-06
139Playground v2 (1024px aesthetic) Playground41±4local~4.9GB2023-12
140Playground v2.5 Playground41±2local~8.8GB2024-02
more from Show Lab, NUS
123Show-o2 1.5B Show Lab, NUS45±5comparison only2025-06
146Show-o2 1.5B-HQ Show Lab, NUS40±3local~7.9GB2025-06
159Show-o (show-o-demo-512) Show Lab, NUS36±3local~7.9GB2024-10

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

$ what it's good at

text to image▂▅▅▃▄▄█▃▁▂▁▁ #84 of 126 · 97% of its score · 4 tests

Beats Latent Diffusion text2img-large (LDM-KL-8-G, LAION-400M), VQ-Diffusion (ITHQ), FLUX.1 Kontext [dev] here, though they rank higher overall. Trails Playground v2 (1024px aesthetic) here, though they rank lower overall.

testresultvs the others in its rundifficultycountswhy
GenExam (relaxed overall, GPT-5 judge)11.2% ✎▄█▂▂▂▂▆▄·▂·▂ beats 3/16hard · avg 26%▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇ 63%sharp
GenEval lab76.0% ✎▇▃▅▇▁▄▅▃█▃▃▅ beats 24/39easy · avg 80%▇▇▇▇▇ 12%at its level · small run
XTC-Bench generation (Overall Gen., scene-graph node agreement)46.5% ✎▄█·▄·······▄ beats 0/4easy · avg 70%▇▇▇▇▇ 12%at its level
DPG-Bench lab86.1% ✎▄·▂·▂▆█▄▂▄▂▂ beats 9/17very easy · avg 88%▇▇▇▇ 10%small run

$ where to run it

Other ways to run it

  1. Run it as published · needs ~22.5 GB, NVIDIA only · from github.com/showlab (official)
    git clone https://github.com/showlab/Show-o && cd Show-o/show-o2 && bash build_env.sh && wandb login && wget https://huggingface.co/Wan-AI/Wan2.1-T2V-14B/resolve/main/Wan2.1_VAE.pth && python3 inference_t2i.py config=configs/showo2_7b_demo_432x432.yaml batch_size=4 guidance_scale=7.5 num_inference_steps=50

Official weights and documentation: huggingface.co/showlab ↗

$ how the score was worked out · GenExam: A Multidisciplinary… 63%, Show Lab Show-o2 paper (arXiv… 22%, Beyond Accuracy: Benchmarking… 12%

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
GenExam: A Multidisciplinary Text-to-Image Exam (arXiv v5)1▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇ 63%CC-BY-4.0
Show Lab Show-o2 paper (arXiv:2506.15564v3), GenEval and DPG-Bench tables (own models) · lab-reported2▇▇▇▇▇▇▇ 22%Factual vendor-reported score (not licensed content)
Beyond Accuracy: Benchmarking Cross-Task Consistency in Unified Multimodal Models (XTC-Bench, arXiv:2604.25072v1)1▇▇▇▇ 12%CC-BY-4.0
prior · pull toward the middle▇ 3%
weighted average= 41.5

$ 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