Instructions to use matu79go/Reflex-1-4B-bnb-4bit with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use matu79go/Reflex-1-4B-bnb-4bit with Transformers:

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

pipe = pipeline("image-text-to-text", model="matu79go/Reflex-1-4B-bnb-4bit")
messages = [
    {
        "role": "user",
        "content": [
            {"type": "image", "url": "https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/p-blog/candy.JPG"},
            {"type": "text", "text": "What animal is on the candy?"}
        ]
    },
]
pipe(text=messages)

# Load model directly
from transformers import AutoProcessor, AutoModelForMultimodalLM

processor = AutoProcessor.from_pretrained("matu79go/Reflex-1-4B-bnb-4bit")
model = AutoModelForMultimodalLM.from_pretrained("matu79go/Reflex-1-4B-bnb-4bit", device_map="auto")
messages = [
    {
        "role": "user",
        "content": [
            {"type": "image", "url": "https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/p-blog/candy.JPG"},
            {"type": "text", "text": "What animal is on the candy?"}
        ]
    },
]
inputs = processor.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(processor.decode(outputs[0][inputs["input_ids"].shape[-1]:]))

Notebooks
Google Colab
Kaggle
Local Apps Settings

vLLM

How to use matu79go/Reflex-1-4B-bnb-4bit with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "matu79go/Reflex-1-4B-bnb-4bit"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "matu79go/Reflex-1-4B-bnb-4bit",
		"messages": [
			{
				"role": "user",
				"content": [
					{
						"type": "text",
						"text": "Describe this image in one sentence."
					},
					{
						"type": "image_url",
						"image_url": {
							"url": "https://cdn.britannica.com/61/93061-050-99147DCE/Statue-of-Liberty-Island-New-York-Bay.jpg"
						}
					}
				]
			}
		]
	}'

Use Docker

docker model run hf.co/matu79go/Reflex-1-4B-bnb-4bit

SGLang

How to use matu79go/Reflex-1-4B-bnb-4bit 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 "matu79go/Reflex-1-4B-bnb-4bit" \
    --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": "matu79go/Reflex-1-4B-bnb-4bit",
		"messages": [
			{
				"role": "user",
				"content": [
					{
						"type": "text",
						"text": "Describe this image in one sentence."
					},
					{
						"type": "image_url",
						"image_url": {
							"url": "https://cdn.britannica.com/61/93061-050-99147DCE/Statue-of-Liberty-Island-New-York-Bay.jpg"
						}
					}
				]
			}
		]
	}'

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 "matu79go/Reflex-1-4B-bnb-4bit" \
        --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": "matu79go/Reflex-1-4B-bnb-4bit",
		"messages": [
			{
				"role": "user",
				"content": [
					{
						"type": "text",
						"text": "Describe this image in one sentence."
					},
					{
						"type": "image_url",
						"image_url": {
							"url": "https://cdn.britannica.com/61/93061-050-99147DCE/Statue-of-Liberty-Island-New-York-Bay.jpg"
						}
					}
				]
			}
		]
	}'

Docker Model Runner
How to use matu79go/Reflex-1-4B-bnb-4bit with Docker Model Runner:
```
docker model run hf.co/matu79go/Reflex-1-4B-bnb-4bit
```

Reflex-1-4B-bnb-4bit

Pre-quantized 4-bit (bitsandbytes NF4, double quantization) copy of Google Gemma 4 E4B for use as the base model of Reflex-1. Download is 9.3 GB instead of about 16 GB, and no quantization is needed at load time.

The vision/audio towers, embeddings and LM head are kept in bfloat16; the transformer blocks are NF4.

Variants

All Reflex-1 repositories: Reflex-1 collection

Repository	What it is	Download	GPU memory
matu79go/Reflex-1-4B	The Reflex-1 skills (LoRA + latent modules) and registry. Needed for every variant	1.8 GB	—
matu79go/Reflex-1-4B-bnb-4bit (this page)	Base model, pre-quantized 4-bit NF4. Recommended	9.3 GB	~10 GB
google/gemma-4-E4B-it	Base model, original BF16 (quantize at load with `--quant nf4`, or run in BF16 with `--quant none`)	~16 GB	~10 GB / ~16 GB

Usage

git clone https://github.com/matu79go/reflex-1.git && cd reflex-1
pip install -r requirements.txt
python -m reflex.server --base matu79go/Reflex-1-4B-bnb-4bit --skills skills.json --port 8097

Results with this base are the same as quantizing google/gemma-4-E4B-it at load time (checked on intent classification, image classification and the three reasoning skills).