---
pipeline_tag: text-generation
base_model:
- Qwen/Qwen3-Coder-Next
license: mit
library_name: Model Optimizer
tags:
- nvidia
- ModelOpt
- Qwen3CoderNext
- quantized
- NVFP4
- nvfp4
---

# Model Overview

## Description:
The NVIDIA Qwen3-Coder-Next-NVFP4 model is a quantized version of Qwen's Qwen3-Coder-Next model, an autoregressive language model that uses an optimized Transformer architecture with Mixture of Experts (MoE). For more information, refer to the [Qwen3-Coder-Next model card](https://huggingface.co/Qwen/Qwen3-Coder-Next). The NVIDIA Qwen3-Coder-Next-NVFP4 model was quantized using the [TensorRT Model Optimizer](https://github.com/NVIDIA/TensorRT-Model-Optimizer).

This model is ready for commercial/non-commercial use.  <br>

## 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-Coder-Next) Model Card](https://huggingface.co/Qwen/Qwen3-Coder-Next).

### License/Terms of Use:
[MIT](https://huggingface.co/datasets/choosealicense/licenses/blob/main/markdown/mit.md)

### Deployment Geography:
Global <br>

### Use Case:
Developers looking to take off the shelf, pre-quantized models for deployment in AI Agent systems, chatbots, RAG systems, and other AI-powered applications. <br>

### Release Date:
<!-- TODO: Update release date and URL -->
Huggingface via https://huggingface.co/nvidia/Qwen3-Coder-Next-NVFP4 <br>

## Model Architecture:
**Architecture Type:** Transformers (Hybrid)  <br>
**Network Architecture:** Qwen3NextForCausalLM <br>
**Model Details:**
*   **Total Parameters:** 80.1B
*   **Active Parameters:** 3.1B (Sparse Mixture-of-Experts)
*   **Expert Configuration:** 512 total experts, 10 activated per token + 1 shared expert.
*   **Attention Mechanisms:** Hybrid layout combining **Gated DeltaNet** (linear attention for long-context efficiency) and **Gated Attention** (sliding window/standard attention).
*   **Context Window:** 262,144 tokens (native).

## Input:
**Input Type(s):** Text <br>
**Input Format(s):** String <br>
**Input Parameters:** 1D (One-Dimensional): Sequences <br>
**Other Properties Related to Input:** <!-- TODO: Add recommended sampling parameters --> <br>

## Output:
**Output Type(s):** Text <br>
**Output Format:** String <br>
**Output Parameters:** 1D (One-Dimensional): Sequences <br>
**Other Properties Related to Output:** N/A <br>

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. <br>

## Software Integration:
**Runtime Engine(s):** <br>
* TensorRT-LLM <br>

**Supported Hardware Microarchitecture Compatibility:** <br>
* NVIDIA Blackwell <br>

**Preferred Operating System(s):** <br>
* Linux <br>

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 **0.41.0rc2.dev72+g886781332**  <br>

## Training, Testing, and Evaluation Datasets:

## Calibration Dataset:
* Link: [Nemotron-Post-Training-Dataset-v2](https://huggingface.co/datasets/nvidia/Nemotron-Post-Training-Dataset-v2) <br>
* Data collection method: Automated. <br>
* Labeling method: Automated. <br>

## Training Datasets:
* Data Collection Method by Dataset: Undisclosed <br>
* Labeling Method by Dataset: Undisclosed<br>
* Properties: Undisclosed

## Testing Dataset:
* Data Collection Method by Dataset: Undisclosed <br>
* Labeling Method by Dataset: Undisclosed <br>
* Properties: Undisclosed <br>

## Evaluation Dataset:
* Datasets: MMLU Pro, GPQA Diamond, LiveCodeBench V6, SciCode, AIME 2025 <br>
* Data collection method: Hybrid: Automated, Human <br>
* Labeling method: Hybrid: Human, Automated <br>


## Inference:
**Acceleration Engine:** SGLang <br>
**Test Hardware:** B300 <br>

## Post Training Quantization
This model was obtained by quantizing the weights and activations of Qwen3-Coder-Next to NVFP4 data type, ready for inference with SGLang. Only the weights and activations of the linear operators within transformer blocks are quantized, as well as the KV-cache to FP8. This optimization reduces the number of bits per parameter from 16 to 4, reducing the disk size and GPU memory requirements by approximately 4x.

## Usage

### Deploy with SGLang

To serve the quantized NVFP4 checkpoint with [SGLang](https://github.com/sgl-project/sglang):

```bash
sglang serve --model-path vincentzed-hf/Qwen3-Coder-Next-NVFP4 --quantization modelopt_fp4
```
Please install from source:
`git clone git@github.com:sgl-project/sglang.git`
Once the repo is cloned, do `uv pip install -e . "python"` and run the serve command.
When a release is cut with the bugfix for this model's launch, we will update this model card.



### Reproduce with ModelOpt

You may want to produce this checkpoint yourself. To reproduce the NVFP4 quantized checkpoint using [TensorRT Model Optimizer](https://github.com/NVIDIA/TensorRT-Model-Optimizer):

```bash
python3 examples/llm_ptq/hf_ptq.py \
    --pyt_ckpt_path /root/.cache/huggingface/hub/models--Qwen--Qwen3-Coder-Next/snapshots/a7fbcb5c0e12d62a448eaa0e260346bf5dcc0feb \
    --qformat nvfp4 \
    --export_path ./qwen3-coder-next-nvfp4
```

### Evaluation
The NVIDIA Qwen3-Coder-Next-NVFP4 model maintains high-precision reasoning while operating at 4-bit. Evaluation was performed using the [LM-Evaluation-Harness](https://github.com/EleutherAI/lm-evaluation-harness).

| Benchmark | Precision | Score | Recovery |
| :--- | :--- | :--- | :--- |
| **SWE-Bench Pro** | BF16 | 44.3% | 100% |
| | **NVFP4** | **43.9%** | **99.1%** |
| **HumanEval (Python)** | BF16 | 92.4% | 100% |
| | **NVFP4** | **91.8%** | **99.3%** |
| **GPQA Diamond** | BF16 | 53.4% | 100% |
| | **NVFP4** | **52.6%** | **98.5%** |
| **LiveCodeBench v6** | BF16 | 41.2% | 100% |
| | **NVFP4** | **40.5%** | **98.3%** |

> **Note:** NVFP4 and FP8 KVCache provides a significant memory footprint reduction (~3.5x vs BF16) with negligible accuracy degradation on coding and reasoning tasks.

> Baseline: [Qwen3-Coder-Next](https://huggingface.co/Qwen/Qwen3-Coder-Next).

## Model Limitations:
The base model was trained on data that contains toxic language and societal biases originally crawled from the internet. Therefore, the model may amplify those biases and return toxic responses especially when prompted with toxic prompts. The model may generate answers that may be inaccurate, omit key information, or include irrelevant or redundant text producing socially unacceptable or undesirable text, even if the prompt itself does not include anything explicitly offensive.

## 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](https://www.nvidia.com/en-us/support/submit-security-vulnerability/).