1. Tutorial Introduction

MiniCPM-V 4.0 is an extremely efficient large-scale end-side model open-sourced by the Natural Language Processing Laboratory of Tsinghua University and Mianbi Intelligence in August 2025. MiniCPM-V 4.0 is built on SigLIP2-400M and MiniCPM4-3B, with a total parameter count of 4.1B. It inherits the powerful single-image, multi-image and video understanding performance of MiniCPM-V 2.6, and greatly improves efficiency. Mianbi Intelligence also simultaneously open-sourced iOS applications for iPhone and iPad. In the OpenCompass evaluation, MiniCPM-V 4.0's image understanding capabilities surpassed GPT-4.1-mini-20250414, Qwen2.5-VL-3B-Instruct and InternVL2.5-8B. The relevant paper results are "MiniCPM-V: A GPT-4V Level MLLM on Your Phone".

This tutorial uses a single RTX 4090 graphics card as computing resource. Three examples are provided for testing: Chat with single or multiple images, Chat with video, and Few shot.

2. Effect display

Deploy MiniCPM-V 4.0 on iPhone 16 Pro Max,iOS DemoThe demo video is a raw, unedited screen recording:

3. Operation steps

1. Start the container

2. Usage steps

If "Bad Gateway" is displayed, it means the model is initializing. Since the model is large, please wait about 2-3 minutes and refresh the page.

Click "How to use" to view the usage guide.

1. Chat with single or multiple images

Specific parameters:

• Decode Type:
  • Beam Search: A highly deterministic decoding method that attempts to retain the most likely candidate sequences. It is more suitable for scenarios that require accurate and consistent output.
  • Sampling: Randomly sampling the next word based on a probability distribution. The output is more creative but may be unstable.

2. Chat with video

3. Few shot

Sample Learning

Specific parameters:

• User: Enter the field to be predicted or learned for this sample.
• Assistant: Enter the value corresponding to the field to be learned for this sample.

Result prediction

4. Discussion

🖌️ If you see a high-quality project, please leave a message in the background to recommend it! In addition, we have also established a tutorial exchange group. Welcome friends to scan the QR code and remark [SD Tutorial] to join the group to discuss various technical issues and share application effects↓

Citation Information

The citation information for this project is as follows:

@article{yao2024minicpm,
  title={MiniCPM-V: A GPT-4V Level MLLM on Your Phone},
  author={Yao, Yuan and Yu, Tianyu and Zhang, Ao and Wang, Chongyi and Cui, Junbo and Zhu, Hongji and Cai, Tianchi and Li, Haoyu and Zhao, Weilin and He, Zhihui and others},
  journal={Nat Commun 16, 5509 (2025)},
  year={2025}
}
@article{yao2024minicpm,
  title={MiniCPM-V: A GPT-4V Level MLLM on Your Phone},
  author={Yao, Yuan and Yu, Tianyu and Zhang, Ao and Wang, Chongyi and Cui, Junbo and Zhu, Hongji and Cai, Tianchi and Li, Haoyu and Zhao, Weilin and He, Zhihui and others},
  journal={arXiv preprint arXiv:2408.01800},
  year={2024}
}