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

Libraries

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

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

pipe = pipeline("text-generation", model="deepseek-ai/DeepSeek-V3.1", trust_remote_code=True)
messages = [
    {"role": "user", "content": "Who are you?"},
]
pipe(messages)

# pip install -U transformers accelerate
# Load model directly
from transformers import AutoTokenizer, AutoModelForCausalLM

tokenizer = AutoTokenizer.from_pretrained("deepseek-ai/DeepSeek-V3.1", trust_remote_code=True)
model = AutoModelForCausalLM.from_pretrained("deepseek-ai/DeepSeek-V3.1", trust_remote_code=True, 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=256)
print(tokenizer.decode(outputs[0][inputs["input_ids"].shape[-1]:]))

Inference
HuggingChat
Notebooks
Google Colab
Kaggle
Local Apps Settings

vLLM

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

Use Docker

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

SGLang

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

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

DeepSeek-V3.1 / assets /search_tool_trajectory.html

msr2000

Release DeepSeek-V3.1

6599f8e about 1 year ago

Raw History Blame Contribute Delete

10.3 kB

	<!doctype html>
	<meta charset="utf-8">
	<style>
	body { margin:16px; font-family:ui-monospace, SFMono-Regular, Menlo, Monaco, Consolas, "Liberation Mono","Courier New", monospace; white-space:pre-wrap; word-break:break-word; line-height:1.6; }
	.token { padding:0 2px; border-radius:4px; }
	.yellow { background:#FFF59D; }
	.blue { background:#BBDEFB; }
	</style>
	<div><span class="token yellow"><｜User｜></span>The current date is 2025-08-16, Saturday. Your primary task is to solve the user's questions, leveraging the appropriate tools as needed.
	[Core Instruction: Language Consistency]
	You MUST write your entire response in the same language as the user's question.
	If the user asks in Chinese, you must think and write in Chinese.
	If the user asks in French, you must think and write in French.
	Do not switch to English unless the user's question was in English. This applies to your reasoning, explanations, and the final answer.
	[Core Instruction: Citation Formatting]
	If the search tool was used and you are referencing a specific webpage in your final answer (the content after <｜search▁end｜>), you must use the format "[citation:x]", where x is the corresponding webpage number.
	You must first engage in step-by-step, rigorous thinking and reasoning before using the appropriate tool or providing the final answer.


	## Tools
	You have access to the following tools:

	### search
	Description: Search the web for information. The result will be a string.

	Parameters: {"type": "object", "properties": {"questions": {"type": "string", "description": "Generate up to 5 search queries based on the requirements below:\n\n- Separate different queries with \"\|\|\". For example: \"Elon Musk lawsuit OpenAI reasons\|\|Elon Musk lawsuit OpenAI latest updates\".\n- All search queries should use generalized, search engine-friendly, and easily retrievable keywords to improve search efficiency and the relevance of results. Avoid directly extracting phrases from the question that are difficult to retrieve effectively. For example: Avoid using relative or ambiguous terms such as “third author” or “the following year.” - Instead, use more universal expressions like “author list” or “award history.”\n- For a single search query, keep them short - remove unnecessary auxiliary words, conjunctions, and interrogative words. Primarily use keyword combinations (separated by spaces). Ideally, they should not exceed 15 characters. Avoid using special punctuation marks.\n- When seeking information from non-English primary sources, adjust search query to retrieve content in the original language (e.g., chinese), even if your initial question is in English. This improves accuracy and avoids translation-related errors.\n- You can search 1 to 5 search queries in parallel each time. If the question involves comparing multiple entities (e.g., people, places, books, movies) or contains two or more sub-questions, please split it into separate queries for each entity or sub-question.\n- Do not re-search for queries that have already been searched."}}, "required": ["questions"]}

	IMPORTANT: ALWAYS adhere to this exact format for tool use:
	<｜tool▁calls▁begin｜><｜tool▁call▁begin｜>tool_call_name<｜tool▁sep｜>tool_call_arguments<｜tool▁call▁end｜>{additional_tool_calls}<｜tool▁calls▁end｜>

	Where:
	- `tool_call_name` must be an exact match to one of the available tools
	- `tool_call_arguments` must be valid JSON that strictly follows the tool's Parameters Schema
	- For multiple tool calls, chain them directly without separators or spaces

	# The user's message is: 有一个全球旗舰店被称为“博物馆”的品牌，在2025年5月初与一家中国度假酒店集团进行的品牌联名，这次联名的宣传片的艺术指导是谁？<span class="token yellow"><｜Assistant｜></span><span class="token yellow"></think></span><span class="token blue"><｜search▁begin｜></span>

	我们将逐步解决这个问题。首先，我们需要找到符合以下条件的品牌：

	1. 全球旗舰店被称为“博物馆”的品牌。
	2. 在2025年5月初与一家中国度假酒店集团进行品牌联名。
	3. 联名宣传片的艺术指导是谁？

	### 第一步：全球旗舰店被称为“博物馆”的品牌

	首先，我们需要找到全球旗舰店被称为“博物馆”的品牌。可能的候选包括：

	[……]

	### 搜索符合条件的品牌

	我将先搜索符合条件的品牌。<span class="token blue"><｜tool▁calls▁begin｜></span><span class="token blue"><｜tool▁call▁begin｜></span>search<span class="token blue"><｜tool▁sep｜></span>{"questions":"全球旗舰店被称为博物馆的品牌\|\|全球旗舰店博物馆\|\|Aesop全球旗舰店博物馆\|\|Aesop旗舰店博物馆\|\|Nike全球旗舰店博物馆"}<span class="token blue"><｜tool▁call▁end｜></span><span class="token blue"><｜tool▁calls▁end｜></span><span class="token blue"><｜end▁of▁sentence｜></span><span class="token yellow"><｜tool▁output▁begin｜></span>[webpage 0 begin]
	[webpage title]始祖鸟在它的“博物馆”里如何讲户外故事？
	[webpage url]https://wap.eastmoney.com/a/202405213082252819.html
	[webpage date published]
	[webpage snippet begin]
	2024年5月21日 — 如何低门槛投资美股？ 5月18日，位于上海南京西路的始祖鸟全球旗舰店正式开业，而这家店的另一个名称是“始祖鸟博物馆”。一楼和二楼则是多种产品线的零售 ...
	[webpage snippet end]
	[webpage content begin]
	[……]
	[webpage content end]
	[webpage 0 end]
	[……]
	<span class="token yellow"><｜tool▁output▁end｜></span>根据搜索结果，符合条件的品牌是始祖鸟（Arc'teryx）。其全球旗舰店被称为“始祖鸟博物馆”，位于上海南京西路。

	### 第二步：品牌联名

	根据问题描述，始祖鸟在2025年5月初与一家中国度假酒店集团进行品牌联名。我们需要找到这次联名的宣传片艺术指导。

	#### 关键点：

	1. 时间：2025年5月初（由于当前时间是2025年8月16日，而联名发生在2025年5月初，因此可能需要搜索相关新闻或公告）。
	2. 联名对象：一家中国度假酒店集团。
	3. 联名内容：品牌联名，宣传片的艺术指导。

	### 第三步：始祖鸟的联名活动

	我需要先找到始祖鸟在2025年5月初与中国度假酒店集团的联名活动。可能的度假酒店集团包括：

	1. 松赞酒店：始祖鸟曾与松赞酒店合作过（根据之前的文章提到）。
	2. 安缦酒店：高端度假酒店。
	3. 悦榕庄：度假酒店集团。
	4. Club Med：度假村集团。
	5. 华住集团：旗下有度假酒店品牌，但可能不符合“度假酒店集团”的描述。
	6. 开元酒店集团：度假酒店品牌。
	7. 柏联酒店：高端度假酒店。

	根据之前的文章提到，始祖鸟曾与松赞酒店合作过联名服饰。因此，松赞酒店可能是联名对象之一。

	### 第四步：搜索联名活动

	我需要搜索始祖鸟在2025年5月初与中国度假酒店集团的联名活动。我将先尝试搜索相关新闻。<span class="token blue"><｜tool▁calls▁begin｜></span><span class="token blue"><｜tool▁call▁begin｜></span>search<span class="token blue"><｜tool▁sep｜></span>{"questions":"始祖鸟2025年5月联名\|\|始祖鸟2025年5月酒店联名\|\|始祖鸟度假酒店联名\|\|始祖鸟宣传片艺术指导\|\|始祖鸟松赞联名2025"}<span class="token blue"><｜tool▁call▁end｜></span><span class="token blue"><｜tool▁calls▁end｜></span><span class="token blue"><｜end▁of▁sentence｜></span><span class="token yellow"><｜tool▁output▁begin｜></span>[webpage 19 begin]
	[webpage title]始祖鸟× 松赞相约纳木错
	[webpage url]https://socialbeta.com/campaign/24810
	[webpage date published]
	[webpage snippet begin]
	始祖鸟与中国高端旅行品牌松赞再次携手，推出以「圣湖」纳木措为灵感的全新联名系列。这已是双方继以两座神山——卡瓦格博与南迦巴瓦为灵感之后的第三次 ...
	[webpage snippet end]
	[webpage content begin]
	[……]
	[webpage content end]
	[webpage 19 end]
	[……]
	<span class="token yellow"><｜tool▁output▁end｜></span>根据搜索结果，始祖鸟在2025年5月初与松赞酒店集团进行了联名合作。这次联名以纳木措湖为灵感，推出了联名系列产品，并拍摄了概念影片《见水知蓝·纳木措》。

	### 第五步：宣传片的艺术指导

	根据搜索结果，宣传片《见水知蓝·纳木措》的艺术指导是国际摄影大师杜可风（Christopher Doyle）。

	### 最终答案

	根据搜索结果，始祖鸟与松赞酒店集团的联名宣传片《见水知蓝·纳木措》的艺术指导是杜可风（Christopher Doyle）。

	<span class="token blue"><｜search▁end｜></span>

	根据搜索结果，始祖鸟（Arc'teryx）在2025年5月初与中国度假酒店集团松赞（Songtsam）进行了品牌联名，推出了以纳木措湖为灵感的联名系列[citation:19][citation:21]。此次联名的宣传片《见水知蓝·纳木措》的艺术指导是国际摄影大师杜可风（Christopher Doyle）[citation:21]。

	### 联名背景：
	- 合作对象：始祖鸟 × 松赞（第三次合作，聚焦纳木措湖）[citation:19][citation:21]
	- 宣传片：《见水知蓝·纳木措》，以纳木措湖的“开湖”时刻为灵感，采用16:9宽幅画面和沉浸式叙事手法，展现高原自然与人文的共生哲学[citation:21]。
	- 艺术指导：杜可风（Christopher Doyle）担任艺术指导，将其标志性的蓝调美学融入影片，强调自然光影和真实记录[citation:21]。

	如需进一步了解联名详情或宣传片内容，可以查阅或[citation:19][citation:21]。</div>