Instructions to use openbmb/MiniCPM-MoE-8x2B with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use openbmb/MiniCPM-MoE-8x2B with Transformers:

# Use a pipeline as a high-level helper
from transformers import pipeline

pipe = pipeline("text-generation", model="openbmb/MiniCPM-MoE-8x2B", trust_remote_code=True)
messages = [
    {"role": "user", "content": "Who are you?"},
]
pipe(messages)

# Load model directly
from transformers import AutoModelForCausalLM
model = AutoModelForCausalLM.from_pretrained("openbmb/MiniCPM-MoE-8x2B", trust_remote_code=True, device_map="auto")

Notebooks
Google Colab
Kaggle
Local Apps Settings

vLLM

How to use openbmb/MiniCPM-MoE-8x2B with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "openbmb/MiniCPM-MoE-8x2B"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "openbmb/MiniCPM-MoE-8x2B",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/openbmb/MiniCPM-MoE-8x2B

SGLang

How to use openbmb/MiniCPM-MoE-8x2B with SGLang:

Install from pip and serve model

# Install SGLang from pip:
pip install sglang
# Start the SGLang server:
python3 -m sglang.launch_server \
    --model-path "openbmb/MiniCPM-MoE-8x2B" \
    --host 0.0.0.0 \
    --port 30000
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:30000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "openbmb/MiniCPM-MoE-8x2B",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker images

docker run --gpus all \
    --shm-size 32g \
    -p 30000:30000 \
    -v ~/.cache/huggingface:/root/.cache/huggingface \
    --env "HF_TOKEN=<secret>" \
    --ipc=host \
    lmsysorg/sglang:latest \
    python3 -m sglang.launch_server \
        --model-path "openbmb/MiniCPM-MoE-8x2B" \
        --host 0.0.0.0 \
        --port 30000
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:30000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "openbmb/MiniCPM-MoE-8x2B",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Docker Model Runner
How to use openbmb/MiniCPM-MoE-8x2B with Docker Model Runner:
```
docker model run hf.co/openbmb/MiniCPM-MoE-8x2B
```

Improve model card

by nielsr HF Staff - opened Jun 10, 2025

base: refs/heads/main

←

from: refs/pr/4

Discussion Files changed

-3

Files changed (1) hide show

README.md +8 -3

README.md CHANGED Viewed

@@ -1,8 +1,14 @@
 # Introduction
 [OpenBMB Technical Blog Series](https://openbmb.vercel.app/)
-The MiniCPM-MoE-8x2B is a decoder-only transformer-based generative language model.
 The MiniCPM-MoE-8x2B adopt a Mixture-of-Experts(MoE) architecture, which has 8 experts per layer and activates 2 of 8 experts for each token.
@@ -30,5 +36,4 @@ print(responds)
 1. As a language model, MiniCPM-MoE-8x2B generates content by learning from a vast amount of text.
 2. However, it does not possess the ability to comprehend or express personal opinions or value judgments.
 3. Any content generated by MiniCPM-MoE-8x2B does not represent the viewpoints or positions of the model developers.
-4. Therefore, when using content generated by MiniCPM-MoE-8x2B, users should take full responsibility for evaluating and verifying it on their own.

+---
+library_name: transformers
+pipeline_tag: text-generation
+license: apache-2.0
+---
 # Introduction
 [OpenBMB Technical Blog Series](https://openbmb.vercel.app/)
+The MiniCPM-MoE-8x2B is a decoder-only transformer-based generative language model. For more details, please refer to our [github repo](https://github.com/OpenBMB/MiniCPM). Also check out the [project page](https://openbmb.vercel.app/?category=Chinese+Blog). The model is described in the paper [MiniCPM4: Ultra-Efficient LLMs on End Devices](https://huggingface.co/papers/2506.07900).
 The MiniCPM-MoE-8x2B adopt a Mixture-of-Experts(MoE) architecture, which has 8 experts per layer and activates 2 of 8 experts for each token.
 1. As a language model, MiniCPM-MoE-8x2B generates content by learning from a vast amount of text.
 2. However, it does not possess the ability to comprehend or express personal opinions or value judgments.
 3. Any content generated by MiniCPM-MoE-8x2B does not represent the viewpoints or positions of the model developers.
+4. Therefore, when using content generated by MiniCPM-MoE-8x2B, users should take full responsibility for evaluating and verifying it on their own.