Qwen
/

Qwen3-32B

Text Generation

text-generation-inference

Model card Files Files and versions

Instructions to use Qwen/Qwen3-32B with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use Qwen/Qwen3-32B with Transformers:

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

pipe = pipeline("text-generation", model="Qwen/Qwen3-32B")
messages = [
    {"role": "user", "content": "Who are you?"},
]
pipe(messages)

# Load model directly
from transformers import AutoTokenizer, AutoModelForCausalLM

tokenizer = AutoTokenizer.from_pretrained("Qwen/Qwen3-32B")
model = AutoModelForCausalLM.from_pretrained("Qwen/Qwen3-32B", 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=40)
print(tokenizer.decode(outputs[0][inputs["input_ids"].shape[-1]:]))

Inference
HuggingChat
Notebooks
Google Colab
Kaggle
AMD Developer Cloud
Local Apps Settings

How to use Qwen/Qwen3-32B with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "Qwen/Qwen3-32B"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "Qwen/Qwen3-32B",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/Qwen/Qwen3-32B

How to use Qwen/Qwen3-32B 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 "Qwen/Qwen3-32B" \
    --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": "Qwen/Qwen3-32B",
		"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 "Qwen/Qwen3-32B" \
        --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": "Qwen/Qwen3-32B",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Docker Model Runner
How to use Qwen/Qwen3-32B with Docker Model Runner:
```
docker model run hf.co/Qwen/Qwen3-32B
```

Resources

View closed (4)

Listed on OpenModelMap

#50 opened 29 days ago by

Qwen3 32B — the teacher for our mobile students

#49 opened about 1 month ago by

Listed on OpenModelMap

#48 opened about 1 month ago by

TemporalMesh Transformer: 29.4 PPL at 48% compute — beats Mamba, new open-source architecture

#47 opened about 2 months ago by

Add LEXam evaluation results

#46 opened 2 months ago by

original_max_position_embeddings for YaRN

#44 opened 6 months ago by

WHEN WILL THE BASE MODEL BE RELEASED? WHEN WILL THE BASE MODEL BE RELEASED? WHEN WILL THE BASE MODEL BE RELEASED?

#43 opened 10 months ago by

When can we see the base model for Qwen3 32B?

#41 opened 12 months ago by

Will Qwen3-32B be updated just like Qwen3-235B-A22B?

#40 opened 12 months ago by

JSON formatting errors in Function Call scenarios

#39 opened about 1 year ago by

qweb1

#38 opened about 1 year ago by

MoE version with the same performance as this 32B dense

#37 opened about 1 year ago by

Low Score on GSM8K on lm-eval-harness? (just 74.91)

#36 opened about 1 year ago by

Transformers version for qwen3-32B-8FP

#35 opened about 1 year ago by

Where is the Base Model?

#34 opened about 1 year ago by

I want to use this model to run my code

#33 opened about 1 year ago by

Qwen3-32B-Base?

#32 opened about 1 year ago by

RuntimeError: probability tensor contains either `inf`, `nan` or element < 0

#31 opened about 1 year ago by

Add assistant mask support to Qwen3-32B

#30 opened about 1 year ago by

Setting 'enable_thinking=False' has no effect.

#29 opened about 1 year ago by

finetune question

#28 opened about 1 year ago by

Qwen3ForCausalLM - Architecture issue

#26 opened about 1 year ago by

Request to Release the Base Model for Qwen3-32B

#25 opened about 1 year ago by

How to control thinking length?

#24 opened about 1 year ago by

Qwen3 does not deploy on Endpoints

#23 opened about 1 year ago by

The model's instructions follow too poorly

#22 opened about 1 year ago by

Update README.md

#21 opened about 1 year ago by

Logical-Transcendence84

please release AWQ version

#20 opened about 1 year ago by

Collections of Bad Cases User Reviews and Comments of Qwen3 32B model

#19 opened over 1 year ago by

Potential issue with large context sizes - can someone confirm?

#18 opened over 1 year ago by

Qwen 3 presence of tools affect output length?

#17 opened over 1 year ago by

"/no_think" control is unstable

#16 opened over 1 year ago by

LICENSE files missing

#14 opened over 1 year ago by

After setting /nothinking or enable_thinking=False, can the empty <thinking> tag be omitted from the response?

#13 opened over 1 year ago by

Feedback: It's a good model, however it hallucinates very badly at local facts (Germany)

#12 opened over 1 year ago by

The correct way of fine-tuning on multi-turn trajectories

#11 opened over 1 year ago by

Providing a GPTQ version

#10 opened over 1 year ago by

how to set, enable_thinking=False, on ollama

#9 opened over 1 year ago by

🚀[Fine-tuning] Implementation and Best Practices for Qwen3 CPT/SFT/DPO/GRPO Training👋

#7 opened over 1 year ago by

Reasoning or Non-reasoning model?

#6 opened over 1 year ago by

Local Installation Video and Testing - Step by Step

#5 opened over 1 year ago by

【Evaluation】Best practice for evaluating Qwen3 !!

#4 opened over 1 year ago by

Base Model?

#3 opened over 1 year ago by

Is this multimodal?

#2 opened over 1 year ago by

Add languages tag

#1 opened over 1 year ago by

de-francophones