deepseek-ai
/

DeepSeek-V3.1-Base

Text Generation

text-generation-inference

Model card Files Files and versions

Instructions to use deepseek-ai/DeepSeek-V3.1-Base with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use deepseek-ai/DeepSeek-V3.1-Base with Transformers:

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

pipe = pipeline("text-generation", model="deepseek-ai/DeepSeek-V3.1-Base", 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-V3.1-Base", trust_remote_code=True)
model = AutoModelForCausalLM.from_pretrained("deepseek-ai/DeepSeek-V3.1-Base", 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]:]))

Notebooks
Google Colab
Kaggle
Local Apps

How to use deepseek-ai/DeepSeek-V3.1-Base with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "deepseek-ai/DeepSeek-V3.1-Base"
# 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-V3.1-Base",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/deepseek-ai/DeepSeek-V3.1-Base

How to use deepseek-ai/DeepSeek-V3.1-Base 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-V3.1-Base" \
    --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-V3.1-Base",
		"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-V3.1-Base" \
        --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-V3.1-Base",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Docker Model Runner
How to use deepseek-ai/DeepSeek-V3.1-Base with Docker Model Runner:
```
docker model run hf.co/deepseek-ai/DeepSeek-V3.1-Base
```

Resources

View closed (8)

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

#60 opened 17 days ago by

Script to download the Base modells tensor files

#59 opened 9 months ago by

the pass@1 of deepseek-v3.1-base in lcb benchmark

#57 opened 9 months ago by

Upload 9c2f9436-6470-49fa-a762-8adb27373e17.jpeg

#56 opened 9 months ago by

Now ! I think it's a best model for world !!!

#55 opened 9 months ago by

🚩 Report: Spam

#54 opened 9 months ago by

《DeepSeek-V3.1思考模式实测：新版本并非全面升级》，欢迎进群交流讨论~

#52 opened 9 months ago by

search-agent是不是无法使用原生的ToolCall？

#51 opened 9 months ago by

You made an error with SimpleQA scores

#50 opened 9 months ago by

火钳刘明 yyds

#49 opened 9 months ago by

The coding capability of v3.1 is truly impressive.

#48 opened 9 months ago by

自言自语会导致胡乱思考

#47 opened 9 months ago by

naked model

#45 opened 9 months ago by

deepseek v3.1在理解用户问题能力/指令遵循上变得有些奇怪

#44 opened 9 months ago by

Will perplexity or other be making a “1776” version of this?

#43 opened 9 months ago by

为什么接在用户消息后的系统消息也会被提到最前面？

#42 opened 9 months ago by

Create README.md

#41 opened 9 months ago by

Create README.md

#40 opened 9 months ago by

前排合照

#39 opened 9 months ago by

火钳留名HYY

#38 opened 9 months ago by

Create README.md

#37 opened 9 months ago by

Create README.md

#36 opened 9 months ago by

ManipulatedConsciousness

Create README.md

#35 opened 9 months ago by

code for game

#32 opened 9 months ago by

did they forget it?

#31 opened 9 months ago by

when can we have the chat model?

#30 opened 9 months ago by

Create README.md

#29 opened 9 months ago by

Create README.md

#28 opened 9 months ago by

hossammogalal999

modelcard

#27 opened 9 months ago by

太新了，火钳刘明

#26 opened 9 months ago by

今天去问了官方客服，这个似乎是融合模型了，期待官方的模型卡

#25 opened 9 months ago by

Create README.md

#24 opened 9 months ago by

Modelcard come On ~~

#23 opened 9 months ago by

Create README.md

#22 opened 9 months ago by

Create README.md

#20 opened 9 months ago by

我要modelcard😡😡😡

#19 opened 9 months ago by

「Feedback」Deepseek V3.1 API 似乎在 JSON Output 上输出空结果的概率提高了

#18 opened 9 months ago by

123

#17 opened 9 months ago by

期待发现更多隐藏能力，加入我们一起评测探讨吧

#16 opened 9 months ago by

This model’s censorship is insane

#15 opened 9 months ago by

Create README.md

#13 opened 9 months ago by

Why taking so long to add more information?

#12 opened 9 months ago by

test

#11 opened 9 months ago by

狗屎一样的更新, 变得和V3老版本一样平庸

#9 opened 9 months ago by

Create README.md

#8 opened 9 months ago by

HunterMefford

#7 opened 9 months ago by

здраствуйте

#6 opened 9 months ago by

keep nucking whale

#5 opened 9 months ago by

The whale is back

#4 opened 9 months ago by

Where can I experience it?

#3 opened 9 months ago by