Instructions to use amd/Qwen3.8-2.4T-A95B-Quark-MXFP4 with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use amd/Qwen3.8-2.4T-A95B-Quark-MXFP4 with Transformers:

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

pipe = pipeline("text-generation", model="amd/Qwen3.8-2.4T-A95B-Quark-MXFP4")
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("amd/Qwen3.8-2.4T-A95B-Quark-MXFP4")
model = AutoModelForCausalLM.from_pretrained("amd/Qwen3.8-2.4T-A95B-Quark-MXFP4", 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 amd/Qwen3.8-2.4T-A95B-Quark-MXFP4 with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "amd/Qwen3.8-2.4T-A95B-Quark-MXFP4"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "amd/Qwen3.8-2.4T-A95B-Quark-MXFP4",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/amd/Qwen3.8-2.4T-A95B-Quark-MXFP4

SGLang

How to use amd/Qwen3.8-2.4T-A95B-Quark-MXFP4 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 "amd/Qwen3.8-2.4T-A95B-Quark-MXFP4" \
    --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": "amd/Qwen3.8-2.4T-A95B-Quark-MXFP4",
		"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 "amd/Qwen3.8-2.4T-A95B-Quark-MXFP4" \
        --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": "amd/Qwen3.8-2.4T-A95B-Quark-MXFP4",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Docker Model Runner
How to use amd/Qwen3.8-2.4T-A95B-Quark-MXFP4 with Docker Model Runner:
```
docker model run hf.co/amd/Qwen3.8-2.4T-A95B-Quark-MXFP4
```

Model Overview

Model Architecture: Qwen3_5MoeForCausalLM
- Input: Text
- Output: Text
Supported Hardware Microarchitecture: AMD MI350 / MI355
ROCm: 7.2.0
PyTorch: 2.12.0
Transformers: 5.14.1
Operating System(s): Linux
Inference Engine: SGLang/vLLM
Model Optimizer: AMD-Quark (v0.12)
- Quantized layers: router experts
- Weight quantization: OCP MXFP4, Static
- Activation quantization: OCP MXFP4, Dynamic

Model Quantization

The model was quantized from Qwen/Qwen3.8-2.4T-A95B-FP8 using AMD-Quark. The weights and activations of the routed experts are quantized to MXFP4.

Quantization scripts:

from quark.torch import LLMTemplate, ModelQuantizer

qwen3_5_moe_text_template = LLMTemplate(
    model_type="qwen3_5_moe_text",
    kv_layers_name=["*k_proj", "*v_proj"],
    q_layer_name="*q_proj"
)
LLMTemplate.register_template(qwen3_5_moe_text_template)

# === NOTE: Change the following source and output dir as needed ===
src_ckpt_path = "Qwen/Qwen3.8-2.4T-A95B-FP8"
output_dir = "amd/Qwen3.8-2.4T-A95B-Quark-MXFP4"

quant_scheme = "mxfp4"
exclude_layers = [
    "lm_head",
    "*self_attn*",
    "*linear_attn*",
    "*mlp.gate",
    "*shared_expert*",
    "mtp.fc",
]
template = LLMTemplate.get("qwen3_5_moe_text")

quant_config = template.get_config(scheme=quant_scheme, exclude_layers=exclude_layers)

# Quantize with File2File mode to avoid loading full model
quantizer = ModelQuantizer(quant_config)
quantizer.direct_quantize_checkpoint(
    pretrained_model_path=src_ckpt_path,
    save_path=output_dir,
)

For further details or issues, please refer to the AMD-Quark documentation or contact the respective developers.

Evaluation

The model was evaluated on the GSM8K benchmark using the SGLang framework. The baseline is the original Qwen/Qwen3.8-2.4T-A95B-FP8 checkpoint.

Accuracy

Benchmark	Qwen/Qwen3.8-2.4T-A95B-FP8	amd/Qwen3.8-2.4T-A95B-Quark-MXFP4 (this model)	Recovery
gsm8k	97.49	97.49	100.00%

Reproduction

The GSM8K results were obtained on SGLang based on the docker image rocm/sgl-dev:v0.5.16-rocm720-mi35x-20260727 using the following recipe.

Serve the model with SGLang:

 python3 -m sglang.launch_server \
    --model-path amd/Qwen3.8-2.4T-A95B-Quark-MXFP4 \
    --served-model-name amd/Qwen3.8-2.4T-A95B-Quark-MXFP4 \
    --tp-size 8 \
    --attention-backend aiter \
    --page-size 1 \
    --kv-cache-dtype auto \
    --chunked-prefill-size 16384 \
    --watchdog-timeout 1200 \
    --mem-fraction-static 0.9 \
    --model-loader-extra-config '{"enable_multithread_load": true}' \
    --trust-remote-code \
    --host 0.0.0.0 \
    --port 9001

Evaluate the model

python3 -m sglang.test.run_eval \
    --port 9001 \
    --model amd/Qwen3.8-2.4T-A95B-Quark-MXFP4 \
    --eval-name gsm8k \
    --num-examples 1319 \
    --num-threads 512 \
    --max-tokens 2048 \
    --chat-template-kwargs '{"enable_thinking": false}'

License

Downloads last month: 9,698

Safetensors

Model size

1.2T params

Tensor type

BF16

Model tree for amd/Qwen3.8-2.4T-A95B-Quark-MXFP4

Base model

Qwen/Qwen3.8-2.4T-A95B

Quantized

Qwen/Qwen3.8-2.4T-A95B-FP8

Quantized

(3)

this model