Instructions to use openbmb/MiniCPM5-2B with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use openbmb/MiniCPM5-2B with Transformers:

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

pipe = pipeline("text-generation", model="openbmb/MiniCPM5-2B")
messages = [
    {"role": "user", "content": "Who are you?"},
]
pipe(messages)

# Load model directly
from transformers import AutoTokenizer, AutoModelForCausalLM

tokenizer = AutoTokenizer.from_pretrained("openbmb/MiniCPM5-2B")
model = AutoModelForCausalLM.from_pretrained("openbmb/MiniCPM5-2B", device_map="auto")
messages = [
    {"role": "user", "content": "Who are you?"},
]
inputs = tokenizer.apply_chat_template(
	messages,
	add_generation_prompt=True,
	tokenize=True,
	return_dict=True,
	return_tensors="pt",
).to(model.device)

outputs = model.generate(**inputs, max_new_tokens=40)
print(tokenizer.decode(outputs[0][inputs["input_ids"].shape[-1]:]))

Notebooks
Google Colab
Kaggle
Local Apps Settings

vLLM

How to use openbmb/MiniCPM5-2B with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "openbmb/MiniCPM5-2B"
# 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/MiniCPM5-2B",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/openbmb/MiniCPM5-2B

SGLang

How to use openbmb/MiniCPM5-2B 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/MiniCPM5-2B" \
    --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/MiniCPM5-2B",
		"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/MiniCPM5-2B" \
        --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/MiniCPM5-2B",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Docker Model Runner
How to use openbmb/MiniCPM5-2B with Docker Model Runner:
```
docker model run hf.co/openbmb/MiniCPM5-2B
```

Tech report link incorrectly points to CPM-4 / 链接错误

by Alice39s - opened about 6 hours ago

Discussion

Alice39s

about 6 hours ago

Hi OpenBMB team,

I noticed that the "tech report" link in the MiniCPM5-2B model card actually redirects to the CPM-4 technical report. Could you please update the link?

你们好，我发现在 MiniCPM5-2B 的模型主页里的 tech report 链接错误跳转到了 MiniCPM-4 的技术报告。

麻烦确认并更新一下链接，谢谢！

funmaker

OpenBMB org about 4 hours ago

Thanks for the reminder!

There is currently no separate technical report for MiniCPM5. The overall training largely follows MiniCPM4's recipe, though we've made some adjustments to the data and post-training pipeline.

If you'd like to dive deeper, the updated README lists the following materials:

Technical report: MiniCPM4 technical report
Core data framework: UltraData – a hierarchical data governance framework (L0–L4), with the accompanying paper UltraData Tiered Data Management
Pre-training data: Ultra-FineWeb, Ultra-FineWeb-L3, and UltraX-Preview
Code and math specific data: UltraData-Code and UltraData-Math
Post-training data: SFT phase: UltraData-SFT-2605 (15M+ samples, including both deep-thinking and non-thinking styles); Agent-specific: UltraData-SFT-Agent-2609; and RL phase: UltraData-RL-2609
RL method: JustRL II – the core idea is to introduce a Critic on top of GRPO for token-level credit assignment, addressing training instability in long CoT scenarios.

Upload images, audio, and videos by dragging in the text input, pasting, or clicking here.

Tap or paste here to upload images

· Sign up or log in to comment