deepseek-ai
/

DeepSeek-R1

Text Generation

text-generation-inference

Model card Files Files and versions

Instructions to use deepseek-ai/DeepSeek-R1 with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use deepseek-ai/DeepSeek-R1 with Transformers:

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

pipe = pipeline("text-generation", model="deepseek-ai/DeepSeek-R1", trust_remote_code=True)
messages = [
    {"role": "user", "content": "Who are you?"},
]
pipe(messages)

# Load model directly
from transformers import AutoTokenizer, AutoModelForCausalLM

tokenizer = AutoTokenizer.from_pretrained("deepseek-ai/DeepSeek-R1", trust_remote_code=True)
model = AutoModelForCausalLM.from_pretrained("deepseek-ai/DeepSeek-R1", trust_remote_code=True)
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]:]))

Inference
HuggingChat
Notebooks
Google Colab
Kaggle
Local Apps

How to use deepseek-ai/DeepSeek-R1 with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "deepseek-ai/DeepSeek-R1"
# 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-R1",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/deepseek-ai/DeepSeek-R1

How to use deepseek-ai/DeepSeek-R1 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-R1" \
    --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-R1",
		"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-R1" \
        --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-R1",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Docker Model Runner
How to use deepseek-ai/DeepSeek-R1 with Docker Model Runner:
```
docker model run hf.co/deepseek-ai/DeepSeek-R1
```

Resources

View closed (39)

Fix `chat_template` crash when assistant message omits the `content` key

#248 opened 9 days ago by

[Feature request] Eliminate pre-attention RMSNorm in MLA-models via scale invariance + weight folding

#247 opened 15 days ago by

DeepSeek-R1-1

#246 opened about 1 month ago by

Install & run deepseek-ai/DeepSeek-R1 easily using llmpm

#245 opened 2 months ago by

Using DeepSeek R1 via API Gateway

#244 opened 3 months ago by

Deepseek R1 Request

#243 opened 3 months ago by

Add MMLU-Pro evaluation result (84.0)

#242 opened 4 months ago by

Add GPQA evaluation result

#241 opened 4 months ago by

Add GPQA evaluation results

#240 opened 4 months ago by

Upload internal.py

#239 opened 4 months ago by

Ananthusajeev190

[Alignment Analysis] R1 hallucinates medical false equivalencies unless strictly constrained (Diabetes vs Psychiatry)

#237 opened 4 months ago by

[Hallucination Report] Model hallucinates biological validity in psychiatric analogies

#236 opened 4 months ago by

Set Max Model Length to correct value

#233 opened 8 months ago by

Update README.md

#232 opened 8 months ago by

Update README.md

#231 opened 9 months ago by

Model's Resource Utilization

#230 opened 9 months ago by

Update README.md

#229 opened 9 months ago by

Tokenizer difference between deepseek and qwen3

#227 opened 11 months ago by

updated version?

#226 opened 12 months ago by

日期出现乱码

#225 opened 12 months ago by

Update README.md

#223 opened about 1 year ago by

dsasd

#222 opened about 1 year ago by

DeepSeek R1 model Bad Cases Genuine User Reviews and Comments Collection

#221 opened about 1 year ago by

Update README.md

#220 opened about 1 year ago by

Update README.md

#219 opened about 1 year ago by

pezhmansamaniii

Update README.md

#218 opened about 1 year ago by

pezhmansamaniii

🚩 Report: Copyright infringement

#217 opened about 1 year ago by

Remove "NOTE: Hugging Face's Transformers has not been directly supported yet." in README

#216 opened about 1 year ago by

Update README.md

#215 opened about 1 year ago by

World's Largest Dataset

#214 opened about 1 year ago by deleted

Request: DOI

#213 opened about 1 year ago by

Knowledge / Training Cutoff of DeepSeek R1

#212 opened about 1 year ago by

Update README.md

#211 opened about 1 year ago by

Make config params float to avoid warning in Transformers

#210 opened about 1 year ago by

Update README.md

#209 opened about 1 year ago by

Update README.md

#207 opened about 1 year ago by

Update README.md

#206 opened about 1 year ago by

DeepSeek中o1-1217的数据是哪里来的。我好像没在OpenAI的官方途径找到，谢谢🙏

#205 opened about 1 year ago by

Any R1 reasoning researchers looking for samples?

#204 opened about 1 year ago by

Update README.md

#203 opened about 1 year ago by

Request: DOI

#202 opened about 1 year ago by

Create 9889555

#201 opened about 1 year ago by

Upload 657f0f06e7ea1b09462a7a16_Feedback and evaluation-p-500.png

#200 opened about 1 year ago by

Best practice for R1 models evaluation: Reasoning efficiency and Performance by MATH-Level

#198 opened about 1 year ago by

DeepSeek R1 full-power version occasionally ends without returning </think>.

#196 opened about 1 year ago by

deepseek满血版偶现结束没有返回 </think>

#195 opened about 1 year ago by

Standing at a flag in Netherlands

#194 opened about 1 year ago by

Delete Config.json

#193 opened about 1 year ago by

Update README.md

#192 opened about 1 year ago by

为助手回答添加 <think>\n> 标签，确保一致性

#191 opened about 1 year ago by