Instructions to use Scale-or-Reason/Qwen2.5-0.5B-ift with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use Scale-or-Reason/Qwen2.5-0.5B-ift with Transformers:

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

pipe = pipeline("text-generation", model="Scale-or-Reason/Qwen2.5-0.5B-ift")
messages = [
    {"role": "user", "content": "Who are you?"},
]
pipe(messages)

# Load model directly
from transformers import AutoTokenizer, AutoModelForCausalLM

tokenizer = AutoTokenizer.from_pretrained("Scale-or-Reason/Qwen2.5-0.5B-ift")
model = AutoModelForCausalLM.from_pretrained("Scale-or-Reason/Qwen2.5-0.5B-ift")
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 Scale-or-Reason/Qwen2.5-0.5B-ift with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "Scale-or-Reason/Qwen2.5-0.5B-ift"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "Scale-or-Reason/Qwen2.5-0.5B-ift",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/Scale-or-Reason/Qwen2.5-0.5B-ift

SGLang

How to use Scale-or-Reason/Qwen2.5-0.5B-ift 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 "Scale-or-Reason/Qwen2.5-0.5B-ift" \
    --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": "Scale-or-Reason/Qwen2.5-0.5B-ift",
		"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 "Scale-or-Reason/Qwen2.5-0.5B-ift" \
        --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": "Scale-or-Reason/Qwen2.5-0.5B-ift",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Docker Model Runner
How to use Scale-or-Reason/Qwen2.5-0.5B-ift with Docker Model Runner:
```
docker model run hf.co/Scale-or-Reason/Qwen2.5-0.5B-ift
```

Qwen2.5-0.5B-ift / README.md

nielsr HF Staff

Improve model card: Add project page link and full paper title

adac47f verified 10 months ago

preview code

Raw

History Blame

5.37 kB

	---
	datasets:
	- When-Does-Reasoning-Matter/general-reasoning-ift-pairs
	- When-Does-Reasoning-Matter/math-reasoning-ift-pairs
	language:
	- en
	library_name: transformers
	pipeline_tag: text-generation
	tags:
	- generated_from_trainer
	---

	# When Does Reasoning Matter?

	<p align="left">
	<img src="https://cdn-avatars.huggingface.co/v1/production/uploads/62be186a5f59ff2320e6e32b/GjJ15tY7-F4bqR96FN4pd.png" alt="Dataset Icon" width="180"/>
	</p>

	<p align="left">
	<a href="https://arxiv.org/pdf/2509.22193" target="_blank" rel="noopener noreferrer">
	<img src="https://img.shields.io/badge/arXiv-2509.22193-b31b1b.svg?style=for-the-badge" alt="arXiv:2509.22193" />
	</a>
	</p>

	Project Page: [https://huggingface.co/when-does-reasoning-matter](https://huggingface.co/when-does-reasoning-matter)

	This model was trained as part of the paper [When Does Reasoning Matter? A Controlled Study of Reasoning's Contribution to Model Performance](https://arxiv.org/pdf/2509.22193)
	It belongs to a collection of General and Math-specific student models distilled from Instruction-Fine-Tuned (IFT) or Reasoning answers generated by [Qwen/Qwen3-235B-A22B](https://huggingface.co/Qwen/Qwen3-235B-A22B).

	<img src="https://huggingface.co/api/resolve-cache/models/When-Does-Reasoning-Matter/Qwen2.5-0.5B-ift/733797fee2fdd300e1a0453d368250327fe4cc44/results.png?%2FWhen-Does-Reasoning-Matter%2FQwen2.5-0.5B-ift%2Fresolve%2Fmain%2Fresults.png=&etag=%22d36dedfbca764a8ac9a7a5ebc043ca53f5ee4966%22" alt="results" width="600"/>

	---

	## Datasets

	These models were trained on the largest set of IFT and Reasoning answer pairs:
	- General dataset: [general-reasoning-ift-pairs](https://huggingface.co/datasets/When-Does-Reasoning-Matter/general-reasoning-ift-pairs)
	- Math dataset: [math-reasoning-ift-pairs](https://huggingface.co/datasets/When-Does-Reasoning-Matter/math-reasoning-ift-pairs)

	---

	## Available Models

	<table>
	<thead>
	<tr>
	<th colspan="2">General</th>
	<th colspan="2">Math</th>
	</tr>
	<tr>
	<th>IFT Models</th>
	<th>Reasoning Models</th>
	<th>IFT Models</th>
	<th>Reasoning Models</th>
	</tr>
	</thead>
	<tbody>
	<tr>
	<td><a href="https://huggingface.co/When-Does-Reasoning-Matter/Qwen2.5-0.5B-ift">Qwen2.5-0.5B-ift</a></td>
	<td><a href="https://huggingface.co/When-Does-Reasoning-Matter/Qwen2.5-0.5B-reasoning">Qwen2.5-0.5B-reasoning</a></td>
	<td><a href="https://huggingface.co/When-Does-Reasoning-Matter/Qwen2.5-0.5B-math-ift">Qwen2.5-0.5B-math-ift</a></td>
	<td><a href="https://huggingface.co/When-Does-Reasoning-Matter/Qwen2.5-0.5B-math-reasoning">Qwen2.5-0.5B-math-reasoning</a></td>
	</tr>
	<tr>
	<td><a href="https://huggingface.co/When-Does-Reasoning-Matter/Qwen2.5-1.5B-ift">Qwen2.5-1.5B-ift</a></td>
	<td><a href="https://huggingface.co/When-Does-Reasoning-Matter/Qwen2.5-1.5B-reasoning">Qwen2.5-1.5B-reasoning</a></td>
	<td><a href="https://huggingface.co/When-Does-Reasoning-Matter/Qwen2.5-1.5B-math-ift">Qwen2.5-1.5B-math-ift</a></td>
	<td><a href="https://huggingface.co/When-Does-Reasoning-Matter/Qwen2.5-1.5B-math-reasoning">Qwen2.5-1.5B-math-reasoning</a></td>
	</tr>
	<tr>
	<td><a href="https://huggingface.co/When-Does-Reasoning-Matter/Qwen2.5-3B-ift">Qwen2.5-3B-ift</a></td>
	<td><a href="https://huggingface.co/When-Does-Reasoning-Matter/Qwen2.5-3B-reasoning">Qwen2.5-3B-reasoning</a></td>
	<td><a href="https://huggingface.co/When-Does-Reasoning-Matter/Qwen2.5-3B-math-ift">Qwen2.5-3B-math-ift</a></td>
	<td><a href="https://huggingface.co/When-Does-Reasoning-Matter/Qwen2.5-3B-math-reasoning">Qwen2.5-3B-math-reasoning</a></td>
	</tr>
	<tr>
	<td><a href="https://huggingface.co/When-Does-Reasoning-Matter/Qwen2.5-7B-ift">Qwen2.5-7B-ift</a></td>
	<td><a href="https://huggingface.co/When-Does-Reasoning-Matter/Qwen2.5-7B-reasoning">Qwen2.5-7B-reasoning</a></td>
	<td><a href="https://huggingface.co/When-Does-Reasoning-Matter/Qwen2.5-7B-math-ift">Qwen2.5-7B-math-ift</a></td>
	<td><a href="https://huggingface.co/When-Does-Reasoning-Matter/Qwen2.5-7B-math-reasoning">Qwen2.5-7B-math-reasoning</a></td>
	</tr>
	<tr>
	<td><a href="https://huggingface.co/When-Does-Reasoning-Matter/Qwen2.5-14B-ift">Qwen2.5-14B-ift</a></td>
	<td><a href="https://huggingface.co/When-Does-Reasoning-Matter/Qwen2.5-14B-reasoning">Qwen2.5-14B-reasoning</a></td>
	<td><a href="https://huggingface.co/When-Does-Reasoning-Matter/Qwen2.5-14B-math-ift">Qwen2.5-14B-math-ift</a></td>
	<td><a href="https://huggingface.co/When-Does-Reasoning-Matter/Qwen2.5-14B-math-reasoning">Qwen2.5-14B-math-reasoning</a></td>
	</tr>
	</tbody>
	</table>

	---

	If you use this dataset in your work, please cite: [When Does Reasoning Matter? A Controlled Study of Reasoning's Contribution to Model Performance](https://arxiv.org/pdf/2509.22193)

	```bibtex
	@misc{boizard2025doesreasoningmattercontrolled,
	title={When Does Reasoning Matter? A Controlled Study of Reasoning's Contribution to Model Performance},
	author={Nicolas Boizard and Hippolyte Gisserot-Boukhlef and Kevin El-Haddad and Céline Hudelot and Pierre Colombo},
	year={2025},
	eprint={2509.22193},
	archivePrefix={arXiv},
	primaryClass={cs.CL},
	url={https://arxiv.org/abs/2509.22193},
	}
	```