Instructions to use wesjos/Qwen3-4B-toolcall with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use wesjos/Qwen3-4B-toolcall with Transformers:

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

pipe = pipeline("text-generation", model="wesjos/Qwen3-4B-toolcall")
messages = [
    {"role": "user", "content": "Who are you?"},
]
pipe(messages)

# Load model directly
from transformers import AutoTokenizer, AutoModelForCausalLM

tokenizer = AutoTokenizer.from_pretrained("wesjos/Qwen3-4B-toolcall")
model = AutoModelForCausalLM.from_pretrained("wesjos/Qwen3-4B-toolcall")
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 Settings

vLLM

How to use wesjos/Qwen3-4B-toolcall with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "wesjos/Qwen3-4B-toolcall"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "wesjos/Qwen3-4B-toolcall",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/wesjos/Qwen3-4B-toolcall

SGLang

How to use wesjos/Qwen3-4B-toolcall 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 "wesjos/Qwen3-4B-toolcall" \
    --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": "wesjos/Qwen3-4B-toolcall",
		"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 "wesjos/Qwen3-4B-toolcall" \
        --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": "wesjos/Qwen3-4B-toolcall",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Unsloth Studio

How to use wesjos/Qwen3-4B-toolcall with Unsloth Studio:

Install Unsloth Studio (macOS, Linux, WSL)

curl -fsSL https://unsloth.ai/install.sh | sh
# Run unsloth studio
unsloth studio -H 0.0.0.0 -p 8888
# Then open http://localhost:8888 in your browser
# Search for wesjos/Qwen3-4B-toolcall to start chatting

Install Unsloth Studio (Windows)

irm https://unsloth.ai/install.ps1 | iex
# Run unsloth studio
unsloth studio -H 0.0.0.0 -p 8888
# Then open http://localhost:8888 in your browser
# Search for wesjos/Qwen3-4B-toolcall to start chatting

Using HuggingFace Spaces for Unsloth

# No setup required
# Open https://huggingface.co/spaces/unsloth/studio in your browser
# Search for wesjos/Qwen3-4B-toolcall to start chatting

Load model with FastModel

pip install unsloth
from unsloth import FastModel
model, tokenizer = FastModel.from_pretrained(
    model_name="wesjos/Qwen3-4B-toolcall",
    max_seq_length=2048,
)

Docker Model Runner
How to use wesjos/Qwen3-4B-toolcall with Docker Model Runner:
```
docker model run hf.co/wesjos/Qwen3-4B-toolcall
```

Model information

Finetine on Model：unsloth/Qwen3-4B-unsloth-bnb-4bit
Dataset: interstellarninja/tool-calls-single-reasoning, mlabonne/FineTome-100k, unsloth/OpenMathReasoning-mini

Eval

some improvements are made after finetuning
gpqa: 0.24 --> 0.333
gsm8k: 0.5 --> 0.68
tool_bench: 0.1667 --> 0.3182

Before finetuning(unsloth/Qwen3-4B-unsloth-bnb-4bit)

+---------+------------+-----------------+---------------+-------+---------+---------+
| Model   | Dataset    | Metric          | Subset        |   Num |   Score | Cat.0   |
+=========+============+=================+===============+=======+=========+=========+
| qwen    | gpqa       | AveragePass@1   | gpqa_extended |    50 |  0.24   | default |
+---------+------------+-----------------+---------------+-------+---------+---------+
| qwen    | gpqa       | AveragePass@1   | gpqa_main     |    50 |  0.26   | default |
+---------+------------+-----------------+---------------+-------+---------+---------+
| qwen    | gpqa       | AveragePass@1   | gpqa_diamond  |    50 |  0.22   | default |
+---------+------------+-----------------+---------------+-------+---------+---------+
| qwen    | gpqa       | AveragePass@1   | OVERALL       |   150 |  0.24   | -       |
+---------+------------+-----------------+---------------+-------+---------+---------+
| qwen    | gsm8k      | AverageAccuracy | main          |    50 |  0.5    | default |
+---------+------------+-----------------+---------------+-------+---------+---------+
| qwen    | tool_bench | Act.EM          | in_domain     |    42 |  0.1667 | default |
+---------+------------+-----------------+---------------+-------+---------+---------+
| qwen    | tool_bench | Act.EM          | out_of_domain |    48 |  0.1667 | default |
+---------+------------+-----------------+---------------+-------+---------+---------+
| qwen    | tool_bench | Act.EM          | OVERALL       |    90 |  0.1667 | -       |
+---------+------------+-----------------+---------------+-------+---------+---------+

After finetuning(This model)

+---------+------------+-----------------+---------------+-------+---------+---------+
| Model   | Dataset    | Metric          | Subset        |   Num |   Score | Cat.0   |
+=========+============+=================+===============+=======+=========+=========+
| model   | gpqa       | AveragePass@1   | gpqa_extended |    50 |  0.26   | default |
+---------+------------+-----------------+---------------+-------+---------+---------+
| model   | gpqa       | AveragePass@1   | gpqa_main     |    50 |  0.36   | default |
+---------+------------+-----------------+---------------+-------+---------+---------+
| model   | gpqa       | AveragePass@1   | gpqa_diamond  |    50 |  0.38   | default |
+---------+------------+-----------------+---------------+-------+---------+---------+
| model   | gpqa       | AveragePass@1   | OVERALL       |   150 |  0.3333 | -       |
+---------+------------+-----------------+---------------+-------+---------+---------+
| model   | gsm8k      | AverageAccuracy | main          |    50 |  0.68   | default |
+---------+------------+-----------------+---------------+-------+---------+---------+
| model   | tool_bench | Act.EM          | in_domain     |    41 |  0.3171 | default |
+---------+------------+-----------------+---------------+-------+---------+---------+
| model   | tool_bench | Act.EM          | out_of_domain |    47 |  0.3191 | default |
+---------+------------+-----------------+---------------+-------+---------+---------+
| model   | tool_bench | Act.EM          | OVERALL       |    88 |  0.3182 | -       |
+---------+------------+-----------------+---------------+-------+---------+---------+

Qwen/Qwen3-4B model

+---------+------------+-----------------+---------------+-------+---------+---------+
| Model   | Dataset    | Metric          | Subset        |   Num |   Score | Cat.0   |
+=========+============+=================+===============+=======+=========+=========+
| model   | gpqa       | AveragePass@1   | gpqa_extended |    50 |    0.32 | default |
+---------+------------+-----------------+---------------+-------+---------+---------+
| model   | gpqa       | AveragePass@1   | gpqa_main     |    50 |    0.22 | default |
+---------+------------+-----------------+---------------+-------+---------+---------+
| model   | gpqa       | AveragePass@1   | gpqa_diamond  |    50 |    0.18 | default |
+---------+------------+-----------------+---------------+-------+---------+---------+
| model   | gpqa       | AveragePass@1   | OVERALL       |   150 |    0.24 | -       |
+---------+------------+-----------------+---------------+-------+---------+---------+
| model   | gsm8k      | AverageAccuracy | main          |    50 |    0.48 | default |
+---------+------------+-----------------+---------------+-------+---------+---------+ 
| model   | tool_bench | Act.EM          | in_domain     |    43 |  0.1628 | default |
+---------+------------+-----------------+---------------+-------+---------+---------+
| model   | tool_bench | Act.EM          | out_of_domain |    47 |  0.1702 | default |
+---------+------------+-----------------+---------------+-------+---------+---------+
| model   | tool_bench | Act.EM          | OVERALL       |    90 |  0.1667 | -       |
+---------+------------+-----------------+---------------+-------+---------+---------+

Downloads last month: 5

Safetensors

Model size

4B params

Tensor type

BF16

Model tree for wesjos/Qwen3-4B-toolcall

Unable to build the model tree, the base model loops to the model itself. Learn more.