Instructions to use WeiboAI/VibeThinker-3B with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use WeiboAI/VibeThinker-3B with Transformers:

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

pipe = pipeline("text-generation", model="WeiboAI/VibeThinker-3B")
messages = [
    {"role": "user", "content": "Who are you?"},
]
pipe(messages)

# Load model directly
from transformers import AutoTokenizer, AutoModelForMultimodalLM

tokenizer = AutoTokenizer.from_pretrained("WeiboAI/VibeThinker-3B")
model = AutoModelForMultimodalLM.from_pretrained("WeiboAI/VibeThinker-3B")
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
Notebooks
Google Colab
Kaggle
Local Apps Settings

vLLM

How to use WeiboAI/VibeThinker-3B with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "WeiboAI/VibeThinker-3B"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "WeiboAI/VibeThinker-3B",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/WeiboAI/VibeThinker-3B

SGLang

How to use WeiboAI/VibeThinker-3B 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 "WeiboAI/VibeThinker-3B" \
    --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": "WeiboAI/VibeThinker-3B",
		"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 "WeiboAI/VibeThinker-3B" \
        --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": "WeiboAI/VibeThinker-3B",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Docker Model Runner
How to use WeiboAI/VibeThinker-3B with Docker Model Runner:
```
docker model run hf.co/WeiboAI/VibeThinker-3B
```

I tested this model

by dpe1 - opened 4 days ago

Discussion

dpe1

4 days ago

•

edited 4 days ago

First of all it doesnt support tools. Second of all its not as good as Qwen 3.5 4B, so I think that it is benchmaxed really hard. The python code it produces for linear algebra routines based on paper pdfs is absolute disaster compared to Qwen 3.5 4B. Also I am using Q8 vibe thinker and Q4 Qwen

lsx666

WeiboAI org 4 days ago

Hi, Fair criticism ~ VibeThinker-3B currently does not support tool use, and our coding focus is mainly competitive-programming / verifiable coding tasks. PDF-to-code workflows, implementing linear algebra routines from papers, tool use, and broader real-world coding are not the main target of this release.

chengxian

4 days ago

writing a html gomoku use q6 gguf, really a S**T

lsx666

WeiboAI org 4 days ago

Sorry it performed poorly in that case. A Q6 GGUF may still lose some capability, especially for longer code generation or tasks that require keeping many UI / game-logic details consistent.
With the unquantized HF weights, we have seen it produce some simple HTML pages even without any specific optimization for frontend/web-app generation. But this release is mainly optimized for verifiable math / contest-style coding, not general web development or polished interactive apps like Gomoku.
We’ll keep improving broader practical coding ability, and quantized inference behavior is something we need to document better too.
Have fun~

Nipah

3 days ago

I think... could someone use this model as a very good code synthesizer?

Dibakar49

2 days ago

Hii

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