Instructions to use RedHatAI/Qwen3.8-Flash-Next-NVFP4 with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use RedHatAI/Qwen3.8-Flash-Next-NVFP4 with Transformers:

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

pipe = pipeline("image-text-to-text", model="RedHatAI/Qwen3.8-Flash-Next-NVFP4")
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("RedHatAI/Qwen3.8-Flash-Next-NVFP4")
model = AutoModelForMultimodalLM.from_pretrained("RedHatAI/Qwen3.8-Flash-Next-NVFP4", 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 RedHatAI/Qwen3.8-Flash-Next-NVFP4 with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "RedHatAI/Qwen3.8-Flash-Next-NVFP4"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "RedHatAI/Qwen3.8-Flash-Next-NVFP4",
		"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/RedHatAI/Qwen3.8-Flash-Next-NVFP4

SGLang

How to use RedHatAI/Qwen3.8-Flash-Next-NVFP4 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 "RedHatAI/Qwen3.8-Flash-Next-NVFP4" \
    --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": "RedHatAI/Qwen3.8-Flash-Next-NVFP4",
		"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 "RedHatAI/Qwen3.8-Flash-Next-NVFP4" \
        --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": "RedHatAI/Qwen3.8-Flash-Next-NVFP4",
		"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 RedHatAI/Qwen3.8-Flash-Next-NVFP4 with Docker Model Runner:
```
docker model run hf.co/RedHatAI/Qwen3.8-Flash-Next-NVFP4
```

RedHatAI/Qwen3.8-Flash-Next-NVFP4

Model Overview

Model Architecture: Qwen4ExpForConditionalGeneration
- Input: Text / Image / Video
- Output: Text
Model Optimizations:
- Weight quantization: FP4
- Activation quantization: FP4
Release Date: 2026-08-27
Version: 1.0
Model Developers: RedHatAI

This model is a quantized version of Qwen/Qwen3.8-Flash-Next. It was evaluated to assess its quality in comparison to the unquantized model.

Model Optimizations

This model was obtained by quantizing the weights and activations of the Mixture-of-Experts (MoE) experts in Qwen/Qwen3.8-Flash-Next to NVFP4 (FP4) data type, ready for inference with vLLM.

This reduces the per-weight precision of the MoE expert parameters from 16 to 4 bits, substantially reducing their memory and disk footprint, while the rest of the model is kept in its original BF16 precision.

Only the weights and activations of the MoE expert linear operators are quantized using LLM Compressor.

Deployment

vLLM Serving

vllm serve RedHatAI/Qwen3.8-Flash-Next-NVFP4 \
  --tensor-parallel-size 4 \
  --enable-auto-tool-choice \
  --tool-call-parser qwen3_coder \
  --reasoning-parser qwen3

Adjust the tensor-parallel size and other hardware-specific settings to your deployment — see the vLLM recipe for Qwen3.8-Flash-Next.

Creation

This model was created by applying LLM Compressor with calibration samples from open-perfectblend (1024 samples).

Evaluation

On average, the RedHatAI checkpoint shows the highest accuracy recovery of 99.1% over the Inferact's 97.6% and NVIDIA's 98.0%, likely due to differences in calibration data and observer implementations. All of the evals were collected through Inspect with vLLM server and single seed evaluations.

Accuracy

Category	Benchmark	Qwen/Qwen3.8-Flash-Next (baseline)	RedHatAI/Qwen3.8-Flash-Next-NVFP4 (this model)	Inferact/Qwen3.8-Flash-Next-NVFP4	nvidia/Qwen3.8-Flash-Next-NVFP4
Reasoning	GPQA Diamond	90.4	92.9	91.4	92.4
Knowledge	MMLU-Pro	88.5	88.2	88.0	88.0
Math	AIME 2026	96.7	100	100	100
Coding & Agents	LiveCodeBench	95.0	95.0	94.0	95.0
	Terminal-Bench 2.1	85.5	86.1	84.5	86.4
	SWE-bench Verified	80.6	79.4	79.8	79.8
	SWE-bench Pro	62.9	63.6	60.9	62.8
	DeepSWE 1.1	66.0	62.3	58.1	56.7

Recovery

Recovery is the quantized model's score as a percentage of the unquantized Qwen/Qwen3.8-Flash-Next score, capped at 100%: min(score / unquantized score × 100, 100).

Category	Benchmark	RedHatAI/Qwen3.8-Flash-Next-NVFP4 (this model)	Inferact/Qwen3.8-Flash-Next-NVFP4	nvidia/Qwen3.8-Flash-Next-NVFP4
Reasoning	GPQA Diamond	100	100	100
Knowledge	MMLU-Pro	99.6	99.4	99.4
Math	AIME 2026	100	100	100
Coding & Agents	LiveCodeBench	100	98.9	100
	Terminal-Bench 2.1	100	98.8	100
	SWE-bench Verified	98.5	99.0	99.1
	SWE-bench Pro	100	96.8	99.9
	DeepSWE 1.1	94.3	88.0	85.9
Average		99.1	97.6	98.0

Downloads last month: 887

Safetensors

Model size

180B params

Tensor type

BF16

I64

Model tree for RedHatAI/Qwen3.8-Flash-Next-NVFP4

Base model

Qwen/Qwen3.8-Flash-Next

Quantized

(363)

this model

RedHatAI
/

Qwen3.8-Flash-Next-NVFP4