amd / Qwen3.5-397B-A17B-NVFP4

huggingface.co
Total runs: 2.9K
24-hour runs: 138
7-day runs: 803
30-day runs: 2.6K
Model's Last Updated: September 09 2026
image-text-to-text

Introduction of Qwen3.5-397B-A17B-NVFP4

Model Details of Qwen3.5-397B-A17B-NVFP4

Model Overview

  • Model Architecture: Qwen3_5MoeForConditionalGeneration
    • Input: Text, Image, Video
    • Output: Text
  • Supported Hardware Microarchitecture: AMD MI300/MI350/MI355 (emulation)
  • ROCm: 7.2.2
  • PyTorch : 2.10.0
  • Transformers : 5.2.0
  • Operating System(s): Linux
  • Inference Engine: vLLM
  • Model Optimizer: AMD-Quark (v0.12)
    • Quantized layers: Experts in language model only
    • Weight quantization: MOE-only, NVFP4, Static
    • Activation quantization: MOE-only, NVFP4, Dynamic

Model Quantization

The model was quantized from Qwen/Qwen3.5-397B-A17B-FP8 using AMD-Quark . The weights and activations are quantized to NVFP4.

Quantization scripts:

exclude_layers="lm_head model.visual.* mtp.* *mlp.gate *shared_expert_gate* *linear_attn.* *self_attn.*
export CUDA_VISIBLE_DEVICES=0,1,2,3,4,5,6,7
export MODEL_DIR=Qwen/Qwen3.5-397B-A17B
export output_dir=amd/Qwen3.5-397B-A17B-NVFP4
python3 quantize_quark.py \
--model_dir $MODEL_DIR \
--quant_scheme nvfp4\
--num_calib_data 128 \
--multi_gpu balanced \
--exclude_layers $exclude_layers \
--model_export hf_format  \
--output_dir $output_dir

Deployment

Use with vLLM

This model can be deployed efficiently using the vLLM backend.

Evaluation

The model was evaluated on GSM8K benchmarks.

Accuracy
Benchmark Qwen/Qwen3.5-397B-A17B-FP8 amd/Qwen3.5-397B-A17B-NVFP4(this model) Recovery
gsm8k (flexible-extract) 95.38 94.84 99.43%
Reproduction

The GSM8K result was obtained using the lm-evaluation-harness framework, based on the Docker image rocm/vllm-dev:nightly_main_20260603 .

Install the lm-eval (Version: 0.4.12) in container first.

pip install lm-eval
pip install lm-eval[api]
Evaluating model in a new terminal
export VLLM_ALLOW_LONG_MAX_MODEL_LEN=1
export CUDA_VISIBLE_DEVICES=0,1,2,3,4,5,6,7
lm_eval   --model vllm   
          --model_args pretrained=shareddata/Qwen/Qwen3.5-397B-A17B-NVFP4,tensor_parallel_size=8,max_model_len=262144,gpu_memory_utilization=0.90,max_gen_toks=2048,trust_remote_code=True,reasoning_parser=qwen3   
          --tasks gsm8k  
          --num_fewshot 5   
          --batch_size auto

License

Modifications Copyright(c) 2026 Advanced Micro Devices, Inc. All rights reserved.

Runs of amd Qwen3.5-397B-A17B-NVFP4 on huggingface.co

2.9K
Total runs
138
24-hour runs
301
3-day runs
803
7-day runs
2.6K
30-day runs

More Information About Qwen3.5-397B-A17B-NVFP4 huggingface.co Model

More Qwen3.5-397B-A17B-NVFP4 license Visit here:

https://choosealicense.com/licenses/apache-2.0

Qwen3.5-397B-A17B-NVFP4 huggingface.co

Qwen3.5-397B-A17B-NVFP4 huggingface.co is an AI model on huggingface.co that provides Qwen3.5-397B-A17B-NVFP4's model effect (), which can be used instantly with this amd Qwen3.5-397B-A17B-NVFP4 model. huggingface.co supports a free trial of the Qwen3.5-397B-A17B-NVFP4 model, and also provides paid use of the Qwen3.5-397B-A17B-NVFP4. Support call Qwen3.5-397B-A17B-NVFP4 model through api, including Node.js, Python, http.

Qwen3.5-397B-A17B-NVFP4 huggingface.co Url

https://huggingface.co/amd/Qwen3.5-397B-A17B-NVFP4

amd Qwen3.5-397B-A17B-NVFP4 online free

Qwen3.5-397B-A17B-NVFP4 huggingface.co is an online trial and call api platform, which integrates Qwen3.5-397B-A17B-NVFP4's modeling effects, including api services, and provides a free online trial of Qwen3.5-397B-A17B-NVFP4, you can try Qwen3.5-397B-A17B-NVFP4 online for free by clicking the link below.

amd Qwen3.5-397B-A17B-NVFP4 online free url in huggingface.co:

https://huggingface.co/amd/Qwen3.5-397B-A17B-NVFP4

Qwen3.5-397B-A17B-NVFP4 install

Qwen3.5-397B-A17B-NVFP4 is an open source model from GitHub that offers a free installation service, and any user can find Qwen3.5-397B-A17B-NVFP4 on GitHub to install. At the same time, huggingface.co provides the effect of Qwen3.5-397B-A17B-NVFP4 install, users can directly use Qwen3.5-397B-A17B-NVFP4 installed effect in huggingface.co for debugging and trial. It also supports api for free installation.

Qwen3.5-397B-A17B-NVFP4 install url in huggingface.co:

https://huggingface.co/amd/Qwen3.5-397B-A17B-NVFP4

Url of Qwen3.5-397B-A17B-NVFP4

Qwen3.5-397B-A17B-NVFP4 huggingface.co Url

Provider of Qwen3.5-397B-A17B-NVFP4 huggingface.co

amd
ORGANIZATIONS

Other API from amd

huggingface.co

Total runs: 234.7K
Run Growth: 145.1K
Growth Rate: 64.60%
Updated:April 14 2026
huggingface.co

Total runs: 116.3K
Run Growth: 6.5K
Growth Rate: 5.43%
Updated:July 17 2026
huggingface.co

Total runs: 91.5K
Run Growth: 9.3K
Growth Rate: 9.42%
Updated:June 19 2026
huggingface.co

Total runs: 52.7K
Run Growth: -12.9K
Growth Rate: -24.62%
Updated:July 01 2026
huggingface.co

Total runs: 21.1K
Run Growth: 8.7K
Growth Rate: 40.11%
Updated:July 27 2026
huggingface.co

Total runs: 11.0K
Run Growth: 918
Growth Rate: 8.34%
Updated:August 12 2025
huggingface.co

Total runs: 6.8K
Run Growth: -7.2K
Growth Rate: -106.22%
Updated:October 09 2024
huggingface.co

Total runs: 2.8K
Run Growth: 850
Growth Rate: 30.36%
Updated:June 19 2026
huggingface.co

Total runs: 1.3K
Run Growth: -1.9K
Growth Rate: -146.41%
Updated:May 27 2026
huggingface.co

Total runs: 1.1K
Run Growth: -30.3K
Growth Rate: -2717.15%
Updated:June 19 2026
huggingface.co

Total runs: 608
Run Growth: 245
Growth Rate: 44.14%
Updated:November 03 2025
huggingface.co

Total runs: 544
Run Growth: -3.7K
Growth Rate: -672.79%
Updated:June 19 2026
huggingface.co

Total runs: 542
Run Growth: -392
Growth Rate: -72.32%
Updated:November 15 2025