MiniCPM-V-4.7-35B-A3B-gguf

This repository hosts the GGUF (llama.cpp) quantized version of MiniCPM-V 4.7 35B-A3B. For the original BF16 weights and the full model card, please refer to openbmb/MiniCPM-V-4.7-35B-A3B.

GGUF files#

Two sets of files are provided. They differ only in the embedded chat template — the tensor weights are byte-identical.

Which one to use#

• Root files — thinking is disabled inside the embedded template and cannot be turned on. This is the recommended default for this model, which is not trained as a reasoning model.
• thinking/ files — the official upstream template is restored, so the model emits a [Start thinking] … [End thinking] trace before answering. Useful when you want to see the reasoning steps; note that it multiplies the token count, so set a larger -n.

One-click (macOS / Windows) — install Llama, the desktop app from the llama.cpp team and Hugging Face (brew install --cask llama-app, or the Windows installer), then choose Use this model → Llama on this page. Llama downloads a build that fits your machine and opens the chat window.

Or from the terminal — -hf pulls the model and the matching mmproj automatically:

llama cli   -hf openbmb/MiniCPM-V-4.7-35B-A3B-gguf:Q4_K_M
llama serve -hf openbmb/MiniCPM-V-4.7-35B-A3B-gguf:Q4_K_M     # WebUI + OpenAI-compatible API on :9931

Q4_K_M needs roughly 32 GB of unified memory; use IQ4_XS on smaller machines.

llama-mtmd-cli and llama-server remain available for development; llama cli / llama serve are the normal entrypoints. To use a local file instead, pick a quant above and pass -m with the shared mmproj:

# default (thinking disabled, no flag needed)
llama cli -m MiniCPM-V-4_7-35B-A3B-Q4_K_M.gguf --mmproj mmproj-model-f16.gguf --image img.jpg -p "Describe."

# thinking enabled
llama cli -m thinking/MiniCPM-V-4_7-35B-A3B-Q4_K_M-thinking.gguf --mmproj mmproj-model-f16.gguf \
          --image img.jpg -p "Describe." -n 1024

llama.cpp always injects enable_thinking into the chat template, so --reasoning off has no effect on the root files; on the thinking/ files it disables the trace. mmproj files are shared by both sets — they carry no chat template.

MiniCPM-V 4.7#

GitHub · On-device results · Evaluation · Examples · Usage · Model weights · DSpark-gguf

MiniCPM-V 4.7 is the latest model in the MiniCPM-V series, available in a 35B-A3B variant, with 35B-A3B being the first MoE model in the MiniCPM-V series. 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:

• 📱 The First Open-Source MoE Model with Lossless, Smooth On-Device Inference. With our Lookahead Expert Scheduling (LES) technique introduced alongside MiniCPM-V 4.7, combined with fine-grained layer-wise quantization and DSpark speculative sampling integration, MiniCPM-V 4.7 35B-A3B becomes the first open-source MoE model to run smoothly on-device without compromising performance, achieving a decoding speed of 13+ tokens/s with just 3 GB of memory on iPhone 17 Pro Max.
• 🔥 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 35B-A3B variant of MiniCPM-V 4.7 advances the performance–efficiency Pareto frontier for MLLMs. In Instruct mode, 35B-A3B achieves a higher overall score in less inference time than Qwen3.6 35B-A3B.
• 🚀 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#

Model weights  ·  Inference  ·  Fine-tuning  ·  Live demo  ·  Local deployment  ·  API  ·  Cookbook

On-Device Capability and Efficiency#

Smooth On-Device MoE Inference without compromising performance#

An MoE model can have far more total parameters than it activates for a single token, but its on-device storage requirements are not limited to those active parameters. Different tokens can select different experts at different layers, while device memory often cannot hold all expert weights. Without a latency constraint, weights can remain on SSD or other storage, be loaded as needed, and be released after computation. This makes large-model inference possible, but repeated loading stalls can make decoding too slow for real-time interaction.

Another approach is to restrict the pool of available experts. For example, fixing an expert subset for an entire prompt can reduce loading frequency, but limits subsequent token-level routing to that subset, potentially excluding experts the model would otherwise select and reducing its capabilities. The central goal is therefore smooth inference without compromising performance: preserving model capability while delivering sufficient throughput under limited memory and bandwidth.

MiniCPM-V 4.7 35B-A3B addresses this challenge through three complementary techniques:

• Lookahead Expert Scheduling (LES) uses cross-layer routing prediction to preload experts likely to be needed by the next layer. Among the experts selected by the actual router, it prioritizes those successfully prefetched and overlaps their computation with loading the remaining required experts. Predictions guide only prefetching and scheduling; computation still follows the model's original token-level routing, and required experts missed by the prediction are loaded and evaluated as usual.
• Fine-grained layer-wise quantization reduces the storage footprint and memory traffic of expert weights while preserving model capability, making expert loading better suited to on-device memory and bandwidth constraints.
• DSpark speculative sampling integration incorporates speculative sampling into the on-device inference process, working with expert scheduling and quantization to improve decoding efficiency.

Together, these techniques make MiniCPM-V 4.7 35B-A3B the first open-source MoE model to run smoothly on-device without compromising performance, achieving 13+ tokens/s on iPhone 17 Pro Max.

On-Device MoE Inference: Before and After Optimization

On-device MoE inference speed comparison, with optimization on the left and without optimization on the right

Evaluation#

Significantly Advancing the Performance–Efficiency Pareto Frontier

The 35B-A3B variant of MiniCPM-V 4.7 advances the performance–efficiency Pareto frontier. In Instruct mode, 35B-A3B reaches a higher overall score in less inference time than Qwen3.6 35B-A3B.

Performance versus inference time for MiniCPM-V 4.7 35B-A3B

Overall Performance (35B-A3B, Instruct)

Vision-Language

Vision-Language performance comparison for MiniCPM-V 4.7 35B-A3B in Instruct mode

Language

Language performance comparison for MiniCPM-V 4.7 35B-A3B in Instruct mode

Examples#

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

| 35B-A3B iPhone 17 Pro Max | 35B-A3B iPad M4

| Image: image | Image: image

Usages#

Inference with Transformers#

Click to expand the Transformers inference examples.

Installation#

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:

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

Copy

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

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

Copy

Load the Model#

import torch
from transformers import AutoModelForImageTextToText, AutoProcessor

model_id = "openbmb/MiniCPM-V-4.7-35B-A3B"

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#

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#

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#

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

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#

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-4.7-35B-A3B --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-4.7-35B-A3B",
    "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-4.7-35B-A3B",
  "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#

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.

Speculative decoding (DSpark)#

Click to show speculative decoding.

Speculative decoding (DSpark): we also release MiniCPM-V-4.7-35B-A3B-DSpark, a DSpark draft model trained for MiniCPM-V-4.7-35B-A3B.

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

Acknowledgements#

Click to view acknowledgements.

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

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.

| Model | Description | Model weights | DSpark heads

| MiniCPM‑V 4.7 35B-A3B | The first open-source MoE model to run smoothly on-device without compromising performance, raising the bar for on-device performance. | 🤗  ModelScope BF16

🤗  ModelScope FP8

🤗  ModelScope int4

🤗  ModelScope AWQ

🤗  ModelScope GPTQ

🤗  ModelScope GGUF

| 🤗  ModelScope BF16

🤗  ModelScope GGUF

| MiniCPM‑V 4.7 1B | Combines stronger capabilities with extremely high efficiency, advancing the performance–efficiency Pareto frontier for multimodal large language models. | 🤗  ModelScope BF16

🤗  ModelScope int4

🤗  ModelScope AWQ

🤗  ModelScope GPTQ

🤗  ModelScope GGUF

| 🤗  ModelScope BF16

🤗  ModelScope GGUF

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.