AIDC-AI/Ovis-U1-3B

Ovis-U1

Building on the foundation of the Ovis series, Ovis-U1 is a 3-billion-parameter unified model that seamlessly integrates multimodal understanding, text-to-image generation, and image editing within a single powerful framework.

The overall architecture of Ovis-U1 (cf. Fig.2 in our report).

📦 Installation

Ovis-U1 has been tested with Python 3.10, Torch 2.4.0, Transformers 4.51.3, and DeepSpeed 0.15.4. For a comprehensive list of package dependencies, please consult the requirements.txt file.

git clone [email protected]:AIDC-AI/Ovis-U1.git
conda create -n ovis-u1 python=3.10 -y
conda activate ovis-u1
cd Ovis-U1
pip install -r requirements.txt
pip install -e .

🛠️ Inference

For multimodal understanding, please run

python test_img_to_txt.py

For text-to-image, please run

python test_txt_to_img.py \
    --height 1024 \
    --width 1024  \
    --steps 50 \
    --seed 42 \
    --txt_cfg 5  

For image editing, please run

python test_img_edit.py \
    --steps 50 \
    --img_cfg 1.5 \
    --txt_cfg 6  

📊 Performance

OpenCompass Multi-modal Academic Benchmarks

Model	Avg	MMB	MMS	MMMU	MathVista	Hallusion	AI2D	OCRBench	MMVet
GPT-4o	75.4	86	70.2	72.9	71.6	57	86.3	82.2	76.9
InternVL2.5-2B	59.9	70.9	54.3	43.2	51.1	42.3	74.9	80.2	62.6
SAIL-VL-2B	61	73.7	56.5	44.1	62.8	45.9	77.4	83.1	44.2
InternVL3-2B	61.1	78	61.1	48.7	57.6	41.9	78.6	83.1	67
Qwen2.5-VL-3B	64.5	76.8	56.3	51.2	61.2	46.6	81.4	82.8	60
Ovis2-2B	65.2	76.9	56.7	45.6	64.1	50.2	82.7	87.3	58.3
SAIL-VL-1.5-2B	67	78.5	62.6	46.4	67	50	83.7	89.1	58.8
Ristretto-3B	67.7	80.2	62.8	51.3	67.6	50.2	84.2	84.7	60.7
Ovis-U1	69.6	77.8	61.3	51.1	69.4	56.3	85.6	88.3	66.7

GenEval

Model	Overall	Single object	Two object	Counting	Colors	Position	Attribute binding
GPT-4o	0.84	0.99	0.92	0.85	0.92	0.75	0.61
BAGEL	0.82	0.99	0.94	0.81	0.88	0.64	0.63
BAGEL 📝	0.88	0.98	0.95	0.84	0.95	0.78	0.77
UniWorld-V1	0.80	0.99	0.93	0.79	0.89	0.49	0.70
UniWorld-V1 📝	0.84	0.98	0.93	0.81	0.89	0.74	0.71
OmniGen	0.68	0.98	0.84	0.66	0.74	0.40	0.43
OmniGen2	0.80	1	0.95	0.64	0.88	0.55	0.76
OmniGen2 📝	0.86	0.99	0.96	0.74	0.98	0.71	0.75
Ovis-U1	0.89	0.98	0.98	0.90	0.92	0.79	0.75

📝 denotes using the rewritten prompts

DPG-Bench

Model	Overall	Global	Entity	Attribute	Relation	Other
BAGEL	85.07	88.94	90.37	91.29	90.82	88.67
UniWorld-V1	81.38	83.64	88.39	88.44	89.27	87.22
OmniGen	81.16	87.90	88.97	88.47	87.95	83.56
OmniGen2	83.57	88.81	88.83	90.18	89.37	90.27
Ovis-U1	83.72	82.37	90.08	88.68	93.35	85.20

ImgEdit-Bench

Model	Overall	Add	Adjust	Extract	Replace	Remove	Background	Style	Hybrid	Action
GPT-4o	4.2	4.61	4.33	2.9	4.35	3.66	4.57	4.93	3.96	4.89
MagicBrush	1.90	2.84	1.58	1.51	1.97	1.58	1.75	2.38	1.62	1.22
Instruct-P2P	1.88	2.45	1.83	1.44	2.01	1.50	1.44	3.55	1.2	1.46
AnyEdit	2.45	3.18	2.95	1.88	2.47	2.23	2.24	2.85	1.56	2.65
UltraEdit	2.7	3.44	2.81	2.13	2.96	1.45	2.83	3.76	1.91	2.98
OmniGen	2.96	3.47	3.04	1.71	2.94	2.43	3.21	4.19	2.24	3.38
Step1X-Edit	3.06	3.88	3.14	1.76	3.40	2.41	3.16	4.63	2.64	2.52
ICEdit	3.05	3.58	3.39	1.73	3.15	2.93	3.08	3.84	2.04	3.68
BAGEL	3.2	3.56	3.31	1.7	3.3	2.62	3.24	4.49	2.38	4.17
UniWorld-V1	3.26	3.82	3.64	2.27	3.47	3.24	2.99	4.21	2.96	2.74
OmniGen2	3.44	3.57	3.06	1.77	3.74	3.2	3.57	4.81	2.52	4.68
Ovis-U1	4.00	4.13	3.62	2.98	4.45	4.06	4.22	4.69	3.45	4.61

GEdit-Bench-EN

Model	Avg	Background Change	Color Alteration	Material Modification	Motion Change	Portrait Beautification	Style Transfer	Subject Addition	Subject Removal	Subject Replacement	Text Modification	Tone Transformation
GPT-4o	7.534	7.205	6.491	6.607	8.096	7.768	6.961	7.622	8.331	8.067	7.427	8.301
AnyEdit	3.212	4.663	4.260	2.537	2.024	3.479	2.032	3.995	3.089	3.180	0.922	5.151
Instruct-Pix2Pix	3.684	3.825	5.182	3.688	3.509	4.339	4.560	3.461	2.031	4.237	0.955	4.733
MagicBrush	4.518	5.637	5.136	5.078	4.513	4.487	4.439	5.252	3.704	4.941	1.384	5.130
OmniGen	5.062	5.281	6.003	5.308	2.916	3.087	4.903	6.628	6.352	5.616	4.519	5.064
Gemini	6.315	6.781	6.369	6.040	6.938	5.591	4.676	7.501	6.447	7.003	5.765	6.350
Step1X-Edit	6.701	6.547	6.545	6.204	6.483	6.787	7.221	6.975	6.512	7.068	6.921	6.448
Doubao	6.754	7.430	7.095	6.339	6.973	6.972	6.767	7.674	6.748	7.447	3.471	7.383
BAGEL	6.519	7.324	6.909	6.381	4.753	4.573	6.150	7.896	7.164	7.021	7.320	6.218
Ovis-U1	6.420	7.486	6.879	6.208	4.790	5.981	6.463	7.491	7.254	7.266	4.482	6.314

📚 Citation

If you find Ovis-U1 useful, please cite our paper:

@inproceedings{wang2025ovisu1,
title={Ovis-U1 Technical Report},
author={Ovis Team},
year={2025}
}

🙏 Acknowledgments

The code is built upon Ovis and FLUX.

📄 License

The project is released under Apache License 2.0 (http://www.apache.org/licenses/LICENSE-2.0, SPDX-License-identifier: Apache-2.0).

🚨 Disclaimer

We used compliance checking algorithms during the training process, to ensure the compliance of the trained model to the best of our ability. Due to complex data and the diversity of language model usage scenarios, we cannot guarantee that the model is completely free of copyright issues or improper content. If you believe anything infringes on your rights or generates improper content, please contact us, and we will promptly address the matter.