The NVIDIA Qwen3-VL-235B-A22B NVFP4 model is the quantized version of Alibaba's Qwen3-VL-235B-A22B model, which is an auto-regressive language model that uses an optimized transformer architecture. For more information, please check here. The NVIDIA Qwen3-VL-235B-A22B FP4 model is quantized with TensorRT Model Optimizer.
This model is ready for commercial/non-commercial use.
Third-Party Community Consideration
This model is not owned or developed by NVIDIA. This model has been developed and built to a third-party’s requirements for this application and use case; see link to Non-NVIDIA (Qwen3-VL-235B-A22B) Model Card.
Developers looking to take off-the-shelf, pre-quantized models for deployment in AI Agent systems, chatbots, RAG systems, and other AI-powered applications.
Architecture Type: Transformers Network Architecture: Qwen3-VL-235B-A22B
**This model was developed based on Qwen3-VL-235B-A22B-Instruct
**Number of model parameters: 2.35*10^11.
Input:
Input Type(s): Multilingual text, images and videos Input Format(s):
Input Parameters: One-Dimensional (1D), Two-Dimensional (2D), Three-Dimensional (3D) Other Properties Related to Input: Context length up to 256K natively
Output:
Output Type(s): Text Output Format: String Output Parameters: One-Dimensional (1D): Sequences Other Properties Related to Output: N/A
Our AI models are designed and/or optimized to run on NVIDIA GPU-accelerated systems. By leveraging NVIDIA’s hardware (e.g. GPU cores) and software frameworks (e.g., CUDA libraries), the model achieves faster training and inference times compared to CPU-only solutions.
The integration of foundation and fine-tuned models into AI systems requires additional testing using use-case-specific data to ensure safe and effective deployment. Following the V-model methodology, iterative testing and validation at both unit and system levels are essential to mitigate risks, meet technical and functional requirements, and ensure compliance with safety and ethical standards before deployment.
Model Version(s):
The model is quantized with nvidia-modelopt v0.4.20
Training, Testing, and Evaluation Datasets:
Calibration Dataset:
** Link: cnn_dailymail
** Data Collection Method by dataset: Automated.
** Labeling Method by dataset: Automated.
** Properties: The CNN / DailyMail Dataset is an English-language dataset containing just over 300k unique news articles as written by journalists at CNN and the Daily Mail.
Training Dataset:
** Data Collection Method by Dataset: Undisclosed
** Labeling Method by Dataset: Undisclosed
** Properties: Undisclosed
Testing Dataset:
** Data Collection Method by Dataset: Undisclosed
** Labeling Method by Dataset: Undisclosed
** Properties: Undisclosed
Evaluation Dataset:
** Datasets: MMLU Pro, GPQA Diamond, LiveCodeBench, SciCode, AIME 2024
** Data collection method by dataset: Hybrid: Automated, Human
** Labeling Method by dataset: Hybrid: Human, Automated
Inference:
Acceleration Engine: vLLM Test Hardware: B200
Post Training Quantization
This model was obtained by quantizing the weights and activations of Qwen3-VL-235B-A22B to NVFP4 data type, ready for inference with vLLM. Only the weights and activations of the linear operators within transformer blocks are quantized. This optimization reduces the number of bits per parameter from 16 to 4, reducing the disk size and GPU memory requirements by approximately 3.3x.
Usage
Deploy with vLLM
To deploy the quantized checkpoint with vLLM vLLM API, follow the sample codes below:
from vllm import LLM, SamplingParams
# 1. Configuration
# We recommend running this model on Blackwell and use Tensor Parallel on all available GPUs.
tensor_parallel_size = 4
model_path = "nvidia/Qwen3-VL-235B-A22B-Instruct-NVFP4"
# 2. Initialize the Engine
llm = LLM(
model=model_path,
tensor_parallel_size=tensor_parallel_size,
dtype="auto",
gpu_memory_utilization=0.95, # Maximize VRAM usage for this huge model
max_model_len=32768, # Limit context if you run OOM (Out of Memory)
)
# 3. Prepare Inputs (Chat Format)
messages = [
{
"role": "user",
"content": [
{
"type": "image_url",
"image_url": {
"url": "https://qianwen-res.oss-cn-beijing.aliyuncs.com/Qwen-VL/assets/demo.jpeg"
}
},
{"type": "text", "text": "Describe this image."},
],
}
]
# 4. Inference
outputs = llm.chat(
messages=messages,
sampling_params=SamplingParams(temperature=0.7, max_tokens=128)
)
# 5. Print Result
for output in outputs:
print(output.outputs[0].text)
Evaluation
The accuracy benchmark results are presented in the table below:
Precision
MMLU Pro
GPQA Diamond
LiveCodeBench
SCICODE
AIME 2024
BF16 (AA Ref)
0.82
0.72
0.50
0.41
0.72
FP4
0.82
0.72
0.51
0.41
0.71
Ethical Considerations
NVIDIA believes Trustworthy AI is a shared responsibility and we have established policies and practices to enable development for a wide array of AI applications. When downloaded or used in accordance with our terms of service, developers should work with their internal model team to ensure this model meets requirements for the relevant industry and use case and addresses unforeseen product misuse.
Please report model quality, risk, security vulnerabilities or NVIDIA AI Concerns here.
