Instructions to use aisingapore/Qwen-SEA-LION-v4-32B-IT with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use aisingapore/Qwen-SEA-LION-v4-32B-IT with Transformers:

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

pipe = pipeline("text-generation", model="aisingapore/Qwen-SEA-LION-v4-32B-IT")
messages = [
    {"role": "user", "content": "Who are you?"},
]
pipe(messages)

# Load model directly
from transformers import AutoTokenizer, AutoModelForCausalLM

tokenizer = AutoTokenizer.from_pretrained("aisingapore/Qwen-SEA-LION-v4-32B-IT")
model = AutoModelForCausalLM.from_pretrained("aisingapore/Qwen-SEA-LION-v4-32B-IT")
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 Settings

vLLM

How to use aisingapore/Qwen-SEA-LION-v4-32B-IT with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "aisingapore/Qwen-SEA-LION-v4-32B-IT"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "aisingapore/Qwen-SEA-LION-v4-32B-IT",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/aisingapore/Qwen-SEA-LION-v4-32B-IT

SGLang

How to use aisingapore/Qwen-SEA-LION-v4-32B-IT 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 "aisingapore/Qwen-SEA-LION-v4-32B-IT" \
    --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": "aisingapore/Qwen-SEA-LION-v4-32B-IT",
		"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 "aisingapore/Qwen-SEA-LION-v4-32B-IT" \
        --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": "aisingapore/Qwen-SEA-LION-v4-32B-IT",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Docker Model Runner
How to use aisingapore/Qwen-SEA-LION-v4-32B-IT with Docker Model Runner:
```
docker model run hf.co/aisingapore/Qwen-SEA-LION-v4-32B-IT
```

Any evaluation done in thinking mode?

by ablueleaf - opened Oct 29, 2025

Discussion

ablueleaf

Oct 29, 2025

I'm seeing that all eval for the Qwen Sealion are only in non-thinking mode, any reason why thinking mode eval was not considered?

Jian-Gang

AI Singapore org Oct 29, 2025

Hello,

Thank you for your interest in SEA-LION! Since we have only trained the model in instruct data for now, we have currently only evaluated Qwen-SEA-LION on non-thinking mode. That said, on the SEA-HELM leaderboard, scores for both the original Qwen3-32B non-thinking and thinking modes are available.

ablueleaf

Oct 31, 2025

Hello,

Thank you for your interest in SEA-LION! Since we have only trained the model in instruct data for now, we have currently only evaluated Qwen-SEA-LION on non-thinking mode. That said, on the SEA-HELM leaderboard, scores for both the original Qwen3-32B non-thinking and thinking modes are available.

Interesting, I thought that the CPT on SEA language tokens would have also helped it to reason with a better understanding for such languages and lead to a better conclusion overall at the end of the reasoning? Or is that not how it works?

Jian-Gang

AI Singapore org Nov 5, 2025

Hello,

For SEA-LION, we not only do CPT, but also post-training. Depending on the type of training, the data used for training is also quite different. This is what I meant when I said that we have only trained the model (referring specifically to the post-training phase) on instruct data.

Thank you for the question!

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