---

license: apache-2.0

pipeline_tag: image-text-to-text

library_name: llama.cpp

tags:

• gguf
• llama.cpp
• quantized
• imatrix
• dspark
• speculative-decoding
• minicpm-v
• multimodal
• On-Device Model
• lightweight

base_model: openbmb/MiniCPM-V-4.7-1B

base_model_relation: quantized

---

> This repository hosts the GGUF (llama.cpp) quantized version of MiniCPM-V 4.7 1B. For the original BF16 weights and the full model card, please refer to openbmb/MiniCPM-V-4.7-1B.

MiniCPM-V 4.7

GitHub · Evaluation · Examples · Usage · Model weights · DSpark-gguf

MiniCPM-V 4.7 is the latest model in the MiniCPM-V series, available in a 1B variant. It builds on the family's single-image, multi-image, and video understanding capabilities, combining stronger foundation capabilities with efficient visual encoding and on-device inference. Notable features of MiniCPM-V 4.7 include:

• 🔥 A New Pareto Frontier in Performance and Efficiency.

As model capabilities continue to advance, inference efficiency has become an important measure of a model's value. On resource-constrained devices, delivering the strongest possible capabilities within a limited resource budget calls for higher capability density. The 1B variant of MiniCPM-V 4.7 advances the performance–efficiency Pareto frontier for MLLMs. In Instruct mode, the 1B variant continues to improve foundation capabilities over MiniCPM-V 4.6 and substantially outperforms Qwen3.5 0.8B in both overall score and inference efficiency.

• 🚀 Efficient Architecture.

Based on the intra-ViT early compression technique from LLaVA-UHD v4, MiniCPM-V 4.7 compresses visual tokens early within the vision encoder, reducing computation in subsequent encoder layers. It also supports mixed 4x/16x visual token compression, allowing developers to balance visual detail and inference cost across tasks.

• 💻 Broad Platform and Chip Support.

MiniCPM-V 4.7 supports iOS, Android, and HarmonyOS, together with open-source on-device deployment support. It also provides day-0 support for chip platforms from Rockchip, Houmo, Intel, AMD, and Arm, extending deployment options across mobile and local computing devices.

• 🛠️ Developer Friendly.

MiniCPM-V 4.7 supports inference frameworks including SGLang, vLLM, llama.cpp, and Ollama, along with fine-tuning frameworks including SWIFT and LLaMA-Factory. It provides GGUF, BNB, AWQ, and GPTQ quantized variants and adds FP8 quantization support. We also release Self-Replay Fine-Tuning (SRFT), a fine-tuning approach that helps developers extend model capabilities at low cost while reducing forgetting of existing capabilities, making further customization easier.

Get started

<p align="center">

<a href="https://github.com/OpenBMB/MiniCPM-o/blob/main/docs/minicpm_v4dot7_en.md#model-weights">Model weights</a> &nbsp;·&nbsp;

<a href="https://github.com/OpenBMB/MiniCPM-o/blob/main/docs/minicpm_v4dot7_en.md#inference-with-transformers">Inference</a> &nbsp;·&nbsp;

<a href="https://github.com/OpenBMB/MiniCPM-o/blob/main/docs/deployment_en.md#training-llama-factory-swift">Fine-tuning</a> &nbsp;·&nbsp;

<a href="https://huggingface.co/spaces/openbmb/MiniCPM-V-4.7-Demo">Live demo</a> &nbsp;·&nbsp;

<a href="https://github.com/OpenBMB/MiniCPM-o/blob/main/docs/minicpm_v4dot7_en.md#on-device-deployment">Local deployment</a> &nbsp;·&nbsp;

<a href="https://github.com/OpenBMB/MiniCPM-o/blob/main/docs/api.md">API</a> &nbsp;·&nbsp;

<a href="https://github.com/OpenSQZ/MiniCPM-V-CookBook">Cookbook</a>

</p>

Evaluation

Significantly Advancing the Performance–Efficiency Pareto Frontier

The 1B variant of MiniCPM-V 4.7 advances the performance–efficiency Pareto frontier. In Instruct mode, 1B substantially outperforms Qwen3.5 0.8B in both overall score and inference efficiency.

<p align="center">

<a href="https://raw.githubusercontent.com/OpenBMB/MiniCPM-o/main/assets/minicpmv4.7/performance_efficiency_frontier.png"><img src="https://raw.githubusercontent.com/OpenBMB/MiniCPM-o/main/assets/minicpmv4.7/performance_efficiency_frontier.png" alt="Performance versus inference time for MiniCPM-V 4.7 1B" width="95%"></a>

</p>

Overall Performance (1B, Instruct)

<p align="center">

<a href="https://raw.githubusercontent.com/OpenBMB/MiniCPM-o/main/assets/minicpmv4.7/overall_performance_1b_instruct.png"><img src="https://raw.githubusercontent.com/OpenBMB/MiniCPM-o/main/assets/minicpmv4.7/overall_performance_1b_instruct.png" alt="Full performance comparison for MiniCPM-V 4.7 1B in Instruct mode" width="90%"></a>

</p>

Examples

MiniCPM-V 4.7 can run on phones and tablets. The clips below are unedited screen recordings captured directly on the devices.

<table align="center">

<tr>

<td align="center"><b>1B</b><br>_{iPhone 17 Pro Max}</td>

</tr>

<tr>

<td align="center"><img src="https://raw.githubusercontent.com/OpenBMB/MiniCPM-o/main/assets/minicpmv4.7/v47_1b_iphone_en.gif" height="520"/></td>

</tr>

</table>

Usages

Inference with Transformers <!-- omit in toc -->

<details>

<summary>Click to expand the Transformers inference examples.</summary>

Installation <!-- omit in toc -->

pip install "transformers[torch]>=5.18.0" torchvision torchcodec

> Note on CUDA compatibility: torchcodec (used for video decoding) may have compatibility issues with certain CUDA versions. For example, torch>=2.11 bundles CUDA 13.1 by default, while environments with CUDA 12.x may encounter errors such as RuntimeError: Could not load libtorchcodec. Two workarounds:

>

> 1. Replace torchcodec with PyAV — supports both image and video inference without CUDA version constraints:

> ```bash

> pip install "transformers[torch]>=5.18.0" torchvision av

> ```

> 2. Pin the CUDA version when installing torch to match your environment (e.g. CUDA 12.8):

> ```bash

> pip install "transformers>=5.18.0" torchvision torchcodec --index-url https://download.pytorch.org/whl/cu128

> ```

Load the Model <!-- omit in toc -->

import torch
from transformers import AutoModelForImageTextToText, AutoProcessor

model_id = "openbmb/MiniCPM-V-4.7-1B"

processor = AutoProcessor.from_pretrained(model_id)
model = AutoModelForImageTextToText.from_pretrained(
    model_id, torch_dtype="auto", device_map="auto"
)

# Flash Attention 2 is recommended for faster inference and lower GPU memory usage,
# especially in multi-image and video scenarios.
# model = AutoModelForImageTextToText.from_pretrained(
#     model_id,
#     torch_dtype=torch.bfloat16,
#     attn_implementation="flash_attention_2",
#     device_map="auto",
# )

Image Inference <!-- omit in toc -->

messages = [
    {
        "role": "user",
        "content": [
            {"type": "image", "url": "https://huggingface.co/datasets/openbmb/DemoCase/resolve/main/refract.png"},
            {"type": "text", "text": "What causes this phenomenon?"},
        ],
    }
]

downsample_mode = "16x"  # Using `downsample_mode="4x"` for Finer Detail

inputs = processor.apply_chat_template(
    messages, tokenize=True, add_generation_prompt=True,
    return_dict=True, return_tensors="pt",
    downsample_mode=downsample_mode,
    max_slice_nums=36,
    enable_thinking=False,
).to(model.device)

generated_ids = model.generate(
    **inputs,
    downsample_mode=downsample_mode,
    max_new_tokens=512,
    # recommended instruct-mode sampling parameters
    do_sample=True,
    temperature=0.7,
    top_p=0.95,
    top_k=100,
    min_p=0.0,
    presence_penalty=0,
    repetition_penalty=1.03,
)
generated_ids_trimmed = [
    out_ids[len(in_ids):] for in_ids, out_ids in zip(inputs.input_ids, generated_ids)
]
output_text = processor.batch_decode(
    generated_ids_trimmed, skip_special_tokens=True, clean_up_tokenization_spaces=False
)
print(output_text[0])

Video Inference <!-- omit in toc -->

messages = [
    {
        "role": "user",
        "content": [
            {"type": "video", "url": "https://huggingface.co/datasets/openbmb/DemoCase/resolve/main/football.mp4"},
            {"type": "text", "text": "Describe this video in detail. Follow the timeline and focus on on-screen text, interface changes, main actions, and scene changes."},
        ],
    }
]

downsample_mode = "16x"  # Using `downsample_mode="4x"` for Finer Detail

inputs = processor.apply_chat_template(
    messages, tokenize=True, add_generation_prompt=True,
    return_dict=True, return_tensors="pt",
    downsample_mode=downsample_mode,
    max_num_frames=128,
    stack_frames=1,
    max_slice_nums=1,
    use_image_id=False,
    enable_thinking=False,
).to(model.device)

generated_ids = model.generate(
    **inputs,
    downsample_mode=downsample_mode,
    max_new_tokens=2048,
    # recommended instruct-mode sampling parameters
    do_sample=True,
    temperature=0.7,
    top_p=0.95,
    top_k=100,
    min_p=0.0,
    presence_penalty=0,
    repetition_penalty=1.03,
)
generated_ids_trimmed = [
    out_ids[len(in_ids):] for in_ids, out_ids in zip(inputs.input_ids, generated_ids)
]
output_text = processor.batch_decode(
    generated_ids_trimmed, skip_special_tokens=True, clean_up_tokenization_spaces=False
)
print(output_text[0])

Advanced Parameters <!-- omit in toc -->

You can customize image/video processing by passing additional parameters to apply_chat_template:

| Parameter | Default | Applies to | Description |

|-----------|---------|------------|-------------|

| downsample_mode | "16x" | Image & Video | Visual token downsampling. "16x" merges tokens for efficiency; "4x" keeps four times as many tokens for finer detail. Must also be passed to generate(). |

| max_slice_nums | 9 | Image & Video | Maximum number of slices when splitting a high-resolution image. Higher values preserve more detail for large images. Recommended: 36 for image, 1 for video. |

| max_num_frames | 128 | Video only | Maximum number of main frames sampled from the video. |

| stack_frames | 1 | Video only | Total sample points per second. 1 = main frame only (no stacking). N (N>1) = 1 main frame + N−1 sub-frames per second; the sub-frames are composited into a grid image and interleaved with main frames. Recommended: 3 or 5. |

| use_image_id | True | Image & Video | Whether to prepend <image_id>N</image_id> tags before each image/frame placeholder. Recommended: True for image, False for video. |

> Note: downsample_mode must be passed to both apply_chat_template (for correct placeholder count) and generate (for the vision encoder). All other parameters only need to be passed to apply_chat_template.

Serving with transformers serve <!-- omit in toc -->

Hugging Face Transformers includes a lightweight OpenAI-compatible server for quick testing and moderate-load deployment.

pip install "transformers[serving]>=5.18.0"

Start the server:

transformers serve openbmb/MiniCPM-V-4.7-1B --port 8000 --host 0.0.0.0

Send a request:

curl -s http://localhost:8000/v1/chat/completions \
  -H 'Content-Type: application/json' \
  -d '{
    "model": "openbmb/MiniCPM-V-4.7-1B",
    "messages": [{
      "role": "user",
      "content": [
        {"type": "image_url", "image_url": {"url": "https://huggingface.co/datasets/openbmb/DemoCase/resolve/main/refract.png"}},
        {"type": "text", "text": "What causes this phenomenon?"}
      ]
    }]
  }'

Tool calling example:

curl -s http://localhost:8000/v1/chat/completions -H 'Content-Type: application/json' -d '{
  "model": "openbmb/MiniCPM-V-4.7-1B",
  "messages": [{"role": "user", "content": [
    {"type": "text", "text": "the weather of Beijing"}
  ]}],
  "tools": [{
    "type": "function",
    "function": {
      "name": "get_weather",
      "description": "Get the current weather for a given location",
      "parameters": {
        "type": "object",
        "properties": {
          "location": {"type": "string", "description": "City name"}
        },
        "required": ["location"]
      }
    }
  }]
}'

Recommended Sampling Parameters <!-- omit in toc -->

We recommend using the following set of sampling parameters for generation in instruct mode: temperature=0.7, top_p=0.95, top_k=100, min_p=0.0, presence_penalty=0, repetition_penalty=1.03.

</details>

Speculative decoding (DSpark) <!-- omit in toc -->

<details>

<summary>Click to show speculative decoding. </summary>

Speculative decoding (DSpark): we also release MiniCPM-V-4.7-1B-DSpark, a DSpark draft model trained for MiniCPM-V-4.7-1B.

</details>

Inference with OpenVINO™

OpenVINO™ is an open-source toolkit for optimizing and deploying AI inference on Intel CPU/GPU/NPU platforms. OpenVINO provides day-0 support for MiniCPM-V-4.7 on Intel platforms with high-performance inference. See the OpenVINO™ Toolkit to get started. You can install and run the example with this notebook.

On-Device Deployment

MiniCPM-V 4.7 supports deployment on iOS, Android, and HarmonyOS, with day-0 support for Rockchip, Houmo, Intel, AMD, and Arm chip platforms. See the edge deployment repository for the apps and source code.

Inference and Fine-Tuning

For supported inference and training frameworks, see Supported Inference and Training Frameworks and API. The MiniCPM-V 4.7 Cookbook covers deployment, fine-tuning, and quantization workflows.

Task guides

• Single-image chat
• Multi-image chat
• Video understanding
• OCR
• PDF parsing
• Visual grounding

Quantization

| Method | Guide |

|:---|:---|

| GGUF | Quantization guide |

| BNB | Quantization guide |

| AWQ | Quantization guide |

| GPTQ | Quantization guide |

| FP8 / INT4 | Quantization guide |

Acknowledgements

<details>

<summary>Click to view acknowledgements.</summary>

We would like to thank the following projects:

• Qwen3.5 and Qwen3.6 for providing the language backbone
• SigLIP2 for providing the vision understanding module
• Transformers

</details>

Model Weights

MiniCPM-V 4.7 is available in 35B-A3B and 1B variants. Each model is paired with its DSpark speculative decoding heads, available in BF16 and GGUF.

<table width="840" style="display:table;width:100%;max-width:840px;table-layout:fixed">

<thead>

<tr>

<th width="18%" align="left" style="width:18%">Model</th>

<th width="32%" align="left" style="width:32%">Description</th>

<th width="27%" align="left" style="width:27%">Model weights</th>

<th width="23%" align="left" style="width:23%">DSpark heads</th>

</tr>

</thead>

<tbody>

<tr>

<td style="white-space:nowrap">MiniCPM&#8209;V&nbsp;4.7<br><strong>35B-A3B</strong></td>

<td width="260" style="width:260px;white-space:normal;overflow-wrap:break-word">The first open-source MoE model<br>to run smoothly on-device without<br>compromising performance, raising<br>the bar for on-device performance.</td>

<td style="white-space:nowrap">

<a href="https://huggingface.co/openbmb/MiniCPM-V-4.7-35B-A3B">🤗</a>&nbsp;&nbsp;<a href="https://modelscope.cn/models/OpenBMB/MiniCPM-V-4.7-35B-A3B" style="display:inline-block;vertical-align:middle"><img src="https://raw.githubusercontent.com/OpenBMB/MiniCPM-o/main/assets/modelscope_logo.png" alt="ModelScope" width="20" height="20" style="display:inline-block;vertical-align:middle;margin:0"></a>&nbsp;<strong>BF16</strong><br>

<a href="https://huggingface.co/openbmb/MiniCPM-V-4.7-35B-A3B-FP8">🤗</a>&nbsp;&nbsp;<a href="https://modelscope.cn/models/OpenBMB/MiniCPM-V-4.7-35B-A3B-FP8" style="display:inline-block;vertical-align:middle"><img src="https://raw.githubusercontent.com/OpenBMB/MiniCPM-o/main/assets/modelscope_logo.png" alt="ModelScope" width="20" height="20" style="display:inline-block;vertical-align:middle;margin:0"></a>&nbsp;<strong>FP8</strong><br>

<a href="https://huggingface.co/openbmb/MiniCPM-V-4.7-35B-A3B-int4">🤗</a>&nbsp;&nbsp;<a href="https://modelscope.cn/models/OpenBMB/MiniCPM-V-4.7-35B-A3B-int4" style="display:inline-block;vertical-align:middle"><img src="https://raw.githubusercontent.com/OpenBMB/MiniCPM-o/main/assets/modelscope_logo.png" alt="ModelScope" width="20" height="20" style="display:inline-block;vertical-align:middle;margin:0"></a>&nbsp;<strong>int4</strong><br>

<a href="https://huggingface.co/openbmb/MiniCPM-V-4.7-35B-A3B-AWQ">🤗</a>&nbsp;&nbsp;<a href="https://modelscope.cn/models/OpenBMB/MiniCPM-V-4.7-35B-A3B-AWQ" style="display:inline-block;vertical-align:middle"><img src="https://raw.githubusercontent.com/OpenBMB/MiniCPM-o/main/assets/modelscope_logo.png" alt="ModelScope" width="20" height="20" style="display:inline-block;vertical-align:middle;margin:0"></a>&nbsp;<strong>AWQ</strong><br>

<a href="https://huggingface.co/openbmb/MiniCPM-V-4.7-35B-A3B-GPTQ">🤗</a>&nbsp;&nbsp;<a href="https://modelscope.cn/models/OpenBMB/MiniCPM-V-4.7-35B-A3B-GPTQ" style="display:inline-block;vertical-align:middle"><img src="https://raw.githubusercontent.com/OpenBMB/MiniCPM-o/main/assets/modelscope_logo.png" alt="ModelScope" width="20" height="20" style="display:inline-block;vertical-align:middle;margin:0"></a>&nbsp;<strong>GPTQ</strong><br>

<a href="https://huggingface.co/openbmb/MiniCPM-V-4.7-35B-A3B-gguf">🤗</a>&nbsp;&nbsp;<a href="https://modelscope.cn/models/OpenBMB/MiniCPM-V-4.7-35B-A3B-gguf" style="display:inline-block;vertical-align:middle"><img src="https://raw.githubusercontent.com/OpenBMB/MiniCPM-o/main/assets/modelscope_logo.png" alt="ModelScope" width="20" height="20" style="display:inline-block;vertical-align:middle;margin:0"></a>&nbsp;<strong>GGUF</strong>

</td>

<td style="white-space:nowrap">

<a href="https://huggingface.co/openbmb/MiniCPM-V-4.7-35B-A3B-DSpark">🤗</a>&nbsp;&nbsp;<a href="https://modelscope.cn/models/OpenBMB/MiniCPM-V-4.7-35B-A3B-DSpark" style="display:inline-block;vertical-align:middle"><img src="https://raw.githubusercontent.com/OpenBMB/MiniCPM-o/main/assets/modelscope_logo.png" alt="ModelScope" width="20" height="20" style="display:inline-block;vertical-align:middle;margin:0"></a>&nbsp;<strong>BF16</strong><br>

<a href="https://huggingface.co/openbmb/MiniCPM-V-4.7-35B-A3B-DSpark-gguf">🤗</a>&nbsp;&nbsp;<a href="https://modelscope.cn/models/OpenBMB/MiniCPM-V-4.7-35B-A3B-DSpark-gguf" style="display:inline-block;vertical-align:middle"><img src="https://raw.githubusercontent.com/OpenBMB/MiniCPM-o/main/assets/modelscope_logo.png" alt="ModelScope" width="20" height="20" style="display:inline-block;vertical-align:middle;margin:0"></a>&nbsp;<strong>GGUF</strong>

</td>

</tr>

<tr>

<td style="white-space:nowrap">MiniCPM&#8209;V&nbsp;4.7<br><strong>1B</strong></td>

<td width="260" style="width:260px;white-space:normal;overflow-wrap:break-word">Combines stronger capabilities<br>with extremely high efficiency,<br>advancing the performance–efficiency<br>Pareto frontier for multimodal<br>large language models.</td>

<td style="white-space:nowrap">

<a href="https://huggingface.co/openbmb/MiniCPM-V-4.7-1B">🤗</a>&nbsp;&nbsp;<a href="https://modelscope.cn/models/OpenBMB/MiniCPM-V-4.7-1B" style="display:inline-block;vertical-align:middle"><img src="https://raw.githubusercontent.com/OpenBMB/MiniCPM-o/main/assets/modelscope_logo.png" alt="ModelScope" width="20" height="20" style="display:inline-block;vertical-align:middle;margin:0"></a>&nbsp;<strong>BF16</strong><br>

<a href="https://huggingface.co/openbmb/MiniCPM-V-4.7-1B-int4">🤗</a>&nbsp;&nbsp;<a href="https://modelscope.cn/models/OpenBMB/MiniCPM-V-4.7-1B-int4" style="display:inline-block;vertical-align:middle"><img src="https://raw.githubusercontent.com/OpenBMB/MiniCPM-o/main/assets/modelscope_logo.png" alt="ModelScope" width="20" height="20" style="display:inline-block;vertical-align:middle;margin:0"></a>&nbsp;<strong>int4</strong><br>

<a href="https://huggingface.co/openbmb/MiniCPM-V-4.7-1B-AWQ">🤗</a>&nbsp;&nbsp;<a href="https://modelscope.cn/models/OpenBMB/MiniCPM-V-4.7-1B-AWQ" style="display:inline-block;vertical-align:middle"><img src="https://raw.githubusercontent.com/OpenBMB/MiniCPM-o/main/assets/modelscope_logo.png" alt="ModelScope" width="20" height="20" style="display:inline-block;vertical-align:middle;margin:0"></a>&nbsp;<strong>AWQ</strong><br>

<a href="https://huggingface.co/openbmb/MiniCPM-V-4.7-1B-GPTQ">🤗</a>&nbsp;&nbsp;<a href="https://modelscope.cn/models/OpenBMB/MiniCPM-V-4.7-1B-GPTQ" style="display:inline-block;vertical-align:middle"><img src="https://raw.githubusercontent.com/OpenBMB/MiniCPM-o/main/assets/modelscope_logo.png" alt="ModelScope" width="20" height="20" style="display:inline-block;vertical-align:middle;margin:0"></a>&nbsp;<strong>GPTQ</strong><br>

<a href="https://huggingface.co/openbmb/MiniCPM-V-4.7-1B-gguf">🤗</a>&nbsp;&nbsp;<a href="https://modelscope.cn/models/OpenBMB/MiniCPM-V-4.7-1B-gguf" style="display:inline-block;vertical-align:middle"><img src="https://raw.githubusercontent.com/OpenBMB/MiniCPM-o/main/assets/modelscope_logo.png" alt="ModelScope" width="20" height="20" style="display:inline-block;vertical-align:middle;margin:0"></a>&nbsp;<strong>GGUF</strong>

</td>

<td style="white-space:nowrap">

<a href="https://huggingface.co/openbmb/MiniCPM-V-4.7-1B-DSpark">🤗</a>&nbsp;&nbsp;<a href="https://modelscope.cn/models/OpenBMB/MiniCPM-V-4.7-1B-DSpark" style="display:inline-block;vertical-align:middle"><img src="https://raw.githubusercontent.com/OpenBMB/MiniCPM-o/main/assets/modelscope_logo.png" alt="ModelScope" width="20" height="20" style="display:inline-block;vertical-align:middle;margin:0"></a>&nbsp;<strong>BF16</strong><br>

<a href="https://huggingface.co/openbmb/MiniCPM-V-4.7-1B-DSpark-gguf">🤗</a>&nbsp;&nbsp;<a href="https://modelscope.cn/models/OpenBMB/MiniCPM-V-4.7-1B-DSpark-gguf" style="display:inline-block;vertical-align:middle"><img src="https://raw.githubusercontent.com/OpenBMB/MiniCPM-o/main/assets/modelscope_logo.png" alt="ModelScope" width="20" height="20" style="display:inline-block;vertical-align:middle;margin:0"></a>&nbsp;<strong>GGUF</strong>

</td>

</tr>

</tbody>

</table>

The 1B int4 variant uses bitsandbytes; AWQ and GPTQ use W4A16 quantization. Quantized variants reduce memory usage; GGUF weights support CPU inference.

Also available on Ollama.