Instructions to use MidTool/Arctic-MidTool-MT-8B with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use MidTool/Arctic-MidTool-MT-8B with Transformers:

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

pipe = pipeline("text-generation", model="MidTool/Arctic-MidTool-MT-8B")
messages = [
    {"role": "user", "content": "Who are you?"},
]
pipe(messages)

# Load model directly
from transformers import AutoTokenizer, AutoModelForCausalLM

tokenizer = AutoTokenizer.from_pretrained("MidTool/Arctic-MidTool-MT-8B")
model = AutoModelForCausalLM.from_pretrained("MidTool/Arctic-MidTool-MT-8B", 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]:]))

Notebooks
Google Colab
Kaggle
Local Apps Settings

vLLM

How to use MidTool/Arctic-MidTool-MT-8B with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "MidTool/Arctic-MidTool-MT-8B"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "MidTool/Arctic-MidTool-MT-8B",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/MidTool/Arctic-MidTool-MT-8B

SGLang

How to use MidTool/Arctic-MidTool-MT-8B 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 "MidTool/Arctic-MidTool-MT-8B" \
    --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": "MidTool/Arctic-MidTool-MT-8B",
		"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 "MidTool/Arctic-MidTool-MT-8B" \
        --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": "MidTool/Arctic-MidTool-MT-8B",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Docker Model Runner
How to use MidTool/Arctic-MidTool-MT-8B with Docker Model Runner:
```
docker model run hf.co/MidTool/Arctic-MidTool-MT-8B
```

You need to agree to share your contact information to access this model

This repository is publicly accessible, but you have to accept the conditions to access its files and content.

Arctic-MidTool-MT-8B

Qwen3-8B-Base mid-trained on MidTool-Mix, a 20.3B-token corpus for agentic tool use.

This is the mid-training checkpoint: a base model with a stronger tool-use prior, not an instruction-tuned assistant. Use it as the starting point for your own tool-use SFT/RL. For a ready-to-use agent, see Arctic-MidTool-RL-8B.

Results

Both rows use the same downstream SFT recipe, so the difference isolates the effect of mid-training.

8B setting	BFCLv3 Overall	τ²-Bench Pass@1	MCP-Universe Score
Qwen3-8B-Base + SFT	47.62	10.43	15.18
Arctic-MidTool-MT-8B + SFT	51.12	14.75	17.82

Usage

from transformers import AutoModelForCausalLM, AutoTokenizer

model_id = "MidTool/Arctic-MidTool-MT-8B"
tok = AutoTokenizer.from_pretrained(model_id)
model = AutoModelForCausalLM.from_pretrained(model_id, dtype="auto", device_map="auto")

Details

See our paper for the full data, training, and evaluation details.

@article{jiang2026midtool,
  title  = {MidTool: Mid-training Data Synthesis for Agentic Tool Use},
  author = {Jiang, Fengqing and Wang, Yite and Liu, Boyi and Wang, Zhaoyang and
            Xu, Canwen and Yao, Zhewei and Poovendran, Radha and He, Yuxiong},
  year   = {2026}
}

Downloads last month: -

Safetensors

Model size

8B params

Tensor type

BF16

Model tree for MidTool/Arctic-MidTool-MT-8B

Base model

Qwen/Qwen3-8B-Base

Finetuned

(507)

this model

Finetunes

1 model