whatmodel.sh / image / Ovis-Image-7B
Ovis-Image-7B
↗ share
+ contribute
Ovis-Image-7B is an open-weights image model from Alibaba (AIDC-AI), released November 2025. Its published size is 7B parameters. Licence: apache-2.0.
Catalogue identity notes
Ovis-Image-7B (HF AIDC-AI/Ovis-Image-7B, now served as ATH-MaaS/Ovis-Image-7B), the only public Ovis-Image checkpoint; card: 7B diffusion model plus Ovis2.5-2B text encoder ('2B+7B'), Apache-2.0, diffusers OvisImagePipeline (diffusers>=0.36.0). Memory estimated from 9B parameters at bf16 plus VAE; no VRAM figure published.
Score within the LMArena text-to-image comparison set only.
$ how it compares
The image models ranked either side of Ovis-Image-7B, out of 197 with a score, then more from Alibaba (AIDC-AI).
| # | model | score | where | ||
|---|---|---|---|---|---|
| 111 | LAFITE (HPS v2 checkpoint) | 48±15 | comparison only | ||
| 112 |  SeedEdit 3.0 | 48±3 | comparison only | ||
| 113 | Ovis-Image-7B | 48±7 | local | ||
| 114 |  FLUX.1 Krea [dev] | 47±3 | api $0.025/image +local | ||
| 115 |  MetaQuery-XL | 47±6 | comparison only | ||
| more from Alibaba (AIDC-AI) | |||||
| 165 | Ovis-U1-3B | 35±6 | comparison only | ||
$ what it's good at
text to image▂▅▅▃▄▄█▃▁▂▁▁ #74 of 126 * · 81% of its score · 1 test
Beats Latent Diffusion text2img-large (LDM-KL-8-G, LAION-400M), VQ-Diffusion (ITHQ) here, though they rank higher overall.
| test | result | vs the others in its run | counts |
|---|---|---|---|
| FEPBench instruction faithfulness | 33.4% ✎ | ▄·█···▄▄▄▄▄▄ beats 0/8 | 81% |
$ where to run it
Other ways to run it
version
- Run it as published · needs ~12.1 GB · from github.com/leejet
sd-cli --diffusion-model models/ovis_image-Q4_0.gguf --vae models/ae.safetensors --llm models/ovis_2.5.safetensors -p "a lovely cat" --cfg-scale 5.0 -v --offload-to-cpu --diffusion-fa
Official weights and documentation: huggingface.co/ATH-MaaS ↗
$ how the score was worked out · Faithful, Enriched, and Preci… 81%
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
| source | results | share of the score |
|---|---|---|
 Faithful, Enriched, and Precise: Benchmarking Natural-Science Illustration Generation by T2I models (FEPBench) | 1 | 81% |
| prior · pull toward the middle | 19% | |
| weighted average | = 47.5 |
$ improve this page
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