Instructions to use OpenFlowLM/Qwen3-4B-Instruct-2507-NPU2 with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use OpenFlowLM/Qwen3-4B-Instruct-2507-NPU2 with Transformers:

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

pipe = pipeline("text-generation", model="OpenFlowLM/Qwen3-4B-Instruct-2507-NPU2")
messages = [
    {"role": "user", "content": "Who are you?"},
]
pipe(messages)

# pip install -U transformers accelerate
# Load model directly
from transformers import AutoTokenizer, AutoModelForCausalLM

tokenizer = AutoTokenizer.from_pretrained("OpenFlowLM/Qwen3-4B-Instruct-2507-NPU2")
model = AutoModelForCausalLM.from_pretrained("OpenFlowLM/Qwen3-4B-Instruct-2507-NPU2", 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=256)
print(tokenizer.decode(outputs[0][inputs["input_ids"].shape[-1]:]))

Notebooks
Google Colab
Kaggle
Local Apps Settings

vLLM

How to use OpenFlowLM/Qwen3-4B-Instruct-2507-NPU2 with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "OpenFlowLM/Qwen3-4B-Instruct-2507-NPU2"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "OpenFlowLM/Qwen3-4B-Instruct-2507-NPU2",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/OpenFlowLM/Qwen3-4B-Instruct-2507-NPU2

SGLang

How to use OpenFlowLM/Qwen3-4B-Instruct-2507-NPU2 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 "OpenFlowLM/Qwen3-4B-Instruct-2507-NPU2" \
    --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": "OpenFlowLM/Qwen3-4B-Instruct-2507-NPU2",
		"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 "OpenFlowLM/Qwen3-4B-Instruct-2507-NPU2" \
        --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": "OpenFlowLM/Qwen3-4B-Instruct-2507-NPU2",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Docker Model Runner
How to use OpenFlowLM/Qwen3-4B-Instruct-2507-NPU2 with Docker Model Runner:
```
docker model run hf.co/OpenFlowLM/Qwen3-4B-Instruct-2507-NPU2
```

Qwen3-4B-Instruct-2507-NPU2-Open

OpenFlowLM Q4NX conversion of Qwen/Qwen3-4B-Instruct-2507 for AMD XDNA NPU inference.

This repository contains a quantized Q4NX port of the model, compiled for the OpenFlowLM (OFLM) runtime. It is not a GGUF file.

Item	Value
Source model	`Qwen/Qwen3-4B-Instruct-2507`
Source GGUF	`Qwen3-4B-Instruct-2507.Q8_0.gguf`
Weights	`model.q4nx` (3.07 GB)
Modality	language
OFLM version	`0.1.0`
Converted	2026-10-01

Reproduce this conversion

This container was produced by exactly this command, so a fine-tune of the same base can be packed the same way:

oflm pack -i /home/atomic-germ/.cache/huggingface/hub/models--mradermacher--Qwen3-4B-Instruct-2507-GGUF/snapshots/b374b9c2b665709d2ef157c4f9e432e8ed9c9745/Qwen3-4B-Instruct-2507.Q8_0.gguf -o /home/atomic-germ/Qwen3-4B-Instruct-2507-NPU2-Open -s Qwen/Qwen3-4B-Instruct-2507

Install and run

This repository works with oflm-add, a small installer that copies the model into the OpenFlowLM user directory and registers the tag. It never modifies the system OpenFlowLM install.

pip install oflm-add or uv tool install oflm-add

uv tool install oflm-add
oflm-add Atomic-Germ/Qwen3-4B-Instruct-2507-NPU2-Open --family qwen3.5 --xclbin-from Qwen3-4B-Instruct-2507-NPU2-Open
OFLM_CONFIG_PATH="$HOME/.config/oflm/model_list.json" OFLM_XCLBIN_PATH="$HOME/.config/oflm" oflm run Qwen3-4B-Instruct-2507-NPU2-Open

Files

File	Description
`model.q4nx`	Quantized weights (Q8_0 / Q4_1 / BF16)
`config.json`	OFLM runtime configuration
`tokenizer.json`	Tokenizer vocabulary
`tokenizer_config.json`	Tokenizer configuration
`chat_template.jinja`	Chat template

Source model card

See the original model card: Qwen/Qwen3-4B-Instruct-2507

Downloads last month: 206

Model tree for OpenFlowLM/Qwen3-4B-Instruct-2507-NPU2

Base model

Qwen/Qwen3-4B-Instruct-2507

Quantized

(334)

this model