Instructions to use xguihugging/Qwen3.8-Flash-Next-4layer with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use xguihugging/Qwen3.8-Flash-Next-4layer with Transformers:

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

pipe = pipeline("image-text-to-text", model="xguihugging/Qwen3.8-Flash-Next-4layer")
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)

# pip install -U transformers accelerate
# Load model directly
from transformers import AutoProcessor, AutoModelForMultimodalLM

processor = AutoProcessor.from_pretrained("xguihugging/Qwen3.8-Flash-Next-4layer")
model = AutoModelForMultimodalLM.from_pretrained("xguihugging/Qwen3.8-Flash-Next-4layer", 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=256)
print(processor.decode(outputs[0][inputs["input_ids"].shape[-1]:]))

Notebooks
Google Colab
Kaggle
Local Apps Settings

vLLM

How to use xguihugging/Qwen3.8-Flash-Next-4layer with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "xguihugging/Qwen3.8-Flash-Next-4layer"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "xguihugging/Qwen3.8-Flash-Next-4layer",
		"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/xguihugging/Qwen3.8-Flash-Next-4layer

SGLang

How to use xguihugging/Qwen3.8-Flash-Next-4layer 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 "xguihugging/Qwen3.8-Flash-Next-4layer" \
    --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": "xguihugging/Qwen3.8-Flash-Next-4layer",
		"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 "xguihugging/Qwen3.8-Flash-Next-4layer" \
        --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": "xguihugging/Qwen3.8-Flash-Next-4layer",
		"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 xguihugging/Qwen3.8-Flash-Next-4layer with Docker Model Runner:
```
docker model run hf.co/xguihugging/Qwen3.8-Flash-Next-4layer
```

Qwen3.8-Flash-Next-4layer (CI slice)

Not a usable model. This is a truncated slice of Qwen/Qwen3.8-Flash-Next for GPU CI in SkyRL. It checks numerical parity between HF, vLLM and the Megatron bridge. It is not meant for inference or fine-tuning.

Text layers 0-3 of the original model (3 GatedDeltaNet + 1 gated attention with the QSA indexer), 512-expert MoE, gated hyper-connections, and the final hyper-connection mixer.
The embedding, LM head and vision tower, copied byte-for-byte from the original weights.
MTP layers removed.
PLE n-gram table replaced with a small random table (ngram_vocab_size_base=2048, rows ~ N(0, 0.0075)) in place of the released ~51B-parameter table. ngram_heads_vocab_sizes and ngram_heads_offsets are recomputed to match, so HF, vLLM and Megatron hash identically.

How it was built

python examples/train/qwen3_8_flash_next/make_slice.py <Qwen3.8-Flash-Next> <dst> --num-layers 4

Built with the SkyRL script examples/train/qwen3_8_flash_next/make_slice.py (default --ple-vocab-base 2048, seed 0).

Usage in SkyRL tests

export SKYRL_QWEN38_4LAYER_PATH=<local path or this repo id>

The tests apply Qwen3.8-specific settings only when the path contains qwen3.8-flash-next (case-insensitive).

License

Same as the original model: Qwen Community License 1.0 (see LICENSE).

Downloads last month: 117

Safetensors

Model size

12B params

Tensor type

BF16

I64

Model tree for xguihugging/Qwen3.8-Flash-Next-4layer

Base model

Qwen/Qwen3.8-Flash-Next

Finetuned

(94)

this model

xguihugging
/

Qwen3.8-Flash-Next-4layer

Qwen3.8-Flash-Next-4layer (CI slice)

Contents

How it was built

Usage in SkyRL tests

License

Model tree for xguihugging/Qwen3.8-Flash-Next-4layer