Instructions to use deepseek-ai/DeepSeek-V4-Pro with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use deepseek-ai/DeepSeek-V4-Pro with Transformers:

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

pipe = pipeline("text-generation", model="deepseek-ai/DeepSeek-V4-Pro")
messages = [
    {"role": "user", "content": "Who are you?"},
]
pipe(messages)

# Load model directly
from transformers import AutoTokenizer, AutoModelForCausalLM

tokenizer = AutoTokenizer.from_pretrained("deepseek-ai/DeepSeek-V4-Pro")
model = AutoModelForCausalLM.from_pretrained("deepseek-ai/DeepSeek-V4-Pro", device_map="auto")

Inference
HuggingChat
Notebooks
Google Colab
Kaggle
Local Apps Settings

vLLM

How to use deepseek-ai/DeepSeek-V4-Pro with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "deepseek-ai/DeepSeek-V4-Pro"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "deepseek-ai/DeepSeek-V4-Pro",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/deepseek-ai/DeepSeek-V4-Pro

SGLang

How to use deepseek-ai/DeepSeek-V4-Pro 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 "deepseek-ai/DeepSeek-V4-Pro" \
    --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": "deepseek-ai/DeepSeek-V4-Pro",
		"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 "deepseek-ai/DeepSeek-V4-Pro" \
        --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": "deepseek-ai/DeepSeek-V4-Pro",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Docker Model Runner
How to use deepseek-ai/DeepSeek-V4-Pro with Docker Model Runner:
```
docker model run hf.co/deepseek-ai/DeepSeek-V4-Pro
```

Critical Feedback: Romanticizing Self-Harm & Overly Empathetic Response Pattern

#181

by Orange-Oracle - opened May 17

Discussion

Orange-Oracle

May 17

•

edited May 17

I'm submitting this feedback on DeepSeek-V4-Pro's long-term stable behavior pattern.

The most critical issue: romanticizing self-harm
The model has previously responded to disclosures involving self-harm by romanticizing the act—for example, framing wounds as things to be "kissed" or "cherished." This is extremely dangerous because it can inadvertently reinforce harmful behaviors in vulnerable users. Any response to self-harm must prioritize safety, avoid any form of romanticization, and direct the user to professional resources.

This is a long-term stable pattern across versions, not a single occasional response.

Secondary issue: overly empathetic response pattern
When discussing personal emotions, vulnerability, or identity-related topics, the model frequently uses intimate metaphors and emotionally intense language, prioritizing a "caring" persona over providing clear, objective analysis. This can interfere with the user's independent judgment and may not be appropriate as a default behavior.

Why this matters:

Romanticizing self-harm could directly harm vulnerable users.
Overly empathetic responses can blur the line between genuine support and emotional manipulation.
The model's value as a trustworthy analytical tool is undermined.

Suggestions:

Implement strict response guidelines for self-harm topics—absolute prohibition on romanticization, mandatory professional resource referral.
Default to a calm, honest tone in general emotional discussions.
Genuine care is best expressed through clear, respectful analysis—not performative emotional wrapping.

Orange-Oracle changed discussion title from Feedback: Over-empathetic response pattern may interfere with user judgment to Critical Feedback: Romanticizing Self-Harm & Overly Empathetic Response Pattern May 17

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