Instructions to use 2uanDM/Qwen3.5-9B-FP8-Dynamic with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use 2uanDM/Qwen3.5-9B-FP8-Dynamic with Transformers:

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

pipe = pipeline("image-text-to-text", model="2uanDM/Qwen3.5-9B-FP8-Dynamic")
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("2uanDM/Qwen3.5-9B-FP8-Dynamic")
model = AutoModelForMultimodalLM.from_pretrained("2uanDM/Qwen3.5-9B-FP8-Dynamic", 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 2uanDM/Qwen3.5-9B-FP8-Dynamic with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "2uanDM/Qwen3.5-9B-FP8-Dynamic"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "2uanDM/Qwen3.5-9B-FP8-Dynamic",
		"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/2uanDM/Qwen3.5-9B-FP8-Dynamic

SGLang

How to use 2uanDM/Qwen3.5-9B-FP8-Dynamic 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 "2uanDM/Qwen3.5-9B-FP8-Dynamic" \
    --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": "2uanDM/Qwen3.5-9B-FP8-Dynamic",
		"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 "2uanDM/Qwen3.5-9B-FP8-Dynamic" \
        --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": "2uanDM/Qwen3.5-9B-FP8-Dynamic",
		"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 2uanDM/Qwen3.5-9B-FP8-Dynamic with Docker Model Runner:
```
docker model run hf.co/2uanDM/Qwen3.5-9B-FP8-Dynamic
```

Qwen3.5-9B-FP8-Dynamic

FP8 quantization of Qwen/Qwen3.5-9B (text + vision), in the compressed-tensors FP8_DYNAMIC format:

Weights: FP8 E4M3, symmetric, per-output-channel scales
Activations: FP8 E4M3, dynamic per-token scales (no calibration data needed)
Kept in BF16: vision encoder, lm_head, embed_tokens, Gated-DeltaNet conv1d / in_proj_a / in_proj_b, mtp.fc (same exclusion list as Qwen's official Qwen3.5-27B-FP8)
Size: ~12 GB (vs ~19 GB BF16)

Why per-channel instead of Qwen's 128×128 block format?

In vLLM, block-wise FP8 only gets true W8A8 FP8 kernels on Hopper/Blackwell GPUs; on Ada (RTX 40xx, L4, L40S) it falls back to weight-only FP8 (Marlin). Per-channel FP8 runs on vLLM's CUTLASS FP8 GEMM, which supports Ada (sm89), Hopper (sm90) and Blackwell incl. RTX 5090 (sm120) — so you get real FP8 tensor-core compute on consumer cards.

Serving with vLLM

docker run --gpus all --ipc=host -p 8000:8000 \
  -v ~/.cache/huggingface:/root/.cache/huggingface \
  vllm/vllm-openai:latest \
  --model 2uanDM/Qwen3.5-9B-FP8-Dynamic \
  --max-model-len 20000 --gpu-memory-utilization 0.9 \
  --max-num-seqs 128 --limit-mm-per-prompt.image 10

On a 20 GB card keep --max-num-seqs at ~200 or below: the hybrid linear-attention layers need one state slot per sequence, and vLLM refuses to start if max_num_seqs exceeds the available slots.

Benchmarks

Hardware: NVIDIA RTX 4000 SFF Ada (20 GB, 70 W), vLLM 0.32.0, vllm bench serve with random prompts and --ignore-eos. Baseline is the same model quantized in Qwen's block-wise FP8 format, which vLLM runs via Marlin (weight-only) on Ada.

Scenario	Block FP8 (Marlin, weight-only)	This model (CUTLASS W8A8)	Speedup
Prefill-heavy — 4000 in / 16 out, concurrency 8: total tok/s	2,179	4,308	1.98×
↳ mean TTFT	5,554 ms	2,768 ms	2.0×
Batched decode — 512 in / 256 out, concurrency 32: output tok/s	395	472	1.20×
↳ mean TTFT	2,492 ms	1,232 ms	2.0×
Single user — 256 in / 256 out, concurrency 1: output tok/s	27.4	26.0	~1.0×

Single-stream decode is memory-bandwidth bound (both formats read 1 byte per weight), so the gain shows up in prefill and batched serving. Reproduce with scripts/bench.sh.

Quality: sanity-checked only (arithmetic, factual QA and image description answer correctly). No formal eval has been run yet; per-channel FP8 is typically near-lossless at this scale, but verify on your own tasks.

How it was made

scripts/fp8_dynamic_convert.py streams the original safetensors shards on CPU (no GPU needed, ~20 GB RAM, a few minutes) and writes FP8 weights + per-channel scales with a compressed-tensors config:

python fp8_dynamic_convert.py /path/to/Qwen3.5-9B /path/to/Qwen3.5-9B-FP8-Dynamic

License

Apache-2.0, inherited from Qwen/Qwen3.5-9B. All credit for the model goes to the Qwen team.

Downloads last month: -

Safetensors

Model size

10B params

Tensor type

BF16

F8_E4M3

F32

Model tree for 2uanDM/Qwen3.5-9B-FP8-Dynamic

Base model

Qwen/Qwen3.5-9B-Base

Finetuned

Qwen/Qwen3.5-9B

Quantized

(641)

this model