---
library_name: mlx
license: other
license_name: nvidia-open-model-license
license_link: https://www.nvidia.com/en-us/agreements/enterprise-software/nvidia-open-model-license/
pipeline_tag: text-generation
language:
- en
- es
- fr
- de
- ja
- it
tags:
- nvidia
- pytorch
- mlx
- quantized
- 4bit
datasets:
- nvidia/Nemotron-Pretraining-Code-v1
- nvidia/Nemotron-CC-v2
- nvidia/Nemotron-Pretraining-SFT-v1
- nvidia/Nemotron-CC-Math-v1
- nvidia/Nemotron-Pretraining-Code-v2
- nvidia/Nemotron-Pretraining-Specialized-v1
- nvidia/Nemotron-CC-v2.1
- nvidia/Nemotron-CC-Code-v1
- nvidia/Nemotron-Pretraining-Dataset-sample
- nvidia/Nemotron-Competitive-Programming-v1
- nvidia/Nemotron-Math-v2
- nvidia/Nemotron-Agentic-v1
- nvidia/Nemotron-Math-Proofs-v1
- nvidia/Nemotron-Instruction-Following-Chat-v1
- nvidia/Nemotron-Science-v1
- nvidia/Nemotron-3-Nano-RL-Training-Blend
track_downloads: true
base_model: nvidia/NVIDIA-Nemotron-3-Nano-30B-A3B-BF16
---

# henrybravo/NVIDIA-Nemotron-3-Nano-30B-A3B-4bit

This model is a clean republish of the verified local 4-bit MLX checkpoint for [nvidia/NVIDIA-Nemotron-3-Nano-30B-A3B-BF16](https://huggingface.co/nvidia/NVIDIA-Nemotron-3-Nano-30B-A3B-BF16).

Source checkpoint:

- original MLX repo: [mlx-community/NVIDIA-Nemotron-3-Nano-30B-A3B-4bit](https://huggingface.co/mlx-community/NVIDIA-Nemotron-3-Nano-30B-A3B-4bit)
- quantization: 4-bit affine, group size 64
- local packaging optimization: removed `.cache/huggingface` before upload so the published repo only contains model assets
- local validation stack: `mlx==0.31.0`, `mlx-lm==0.31.0`, `transformers==5.3.0`

This was not re-quantized from the 8-bit checkpoint. Re-quantizing an already quantized model is unnecessary and risks avoidable quality loss.

## Use with mlx

```bash
pip install -U mlx==0.31.0 mlx-lm==0.31.0
```

For OpenAI-compatible local serving and routing on top of MLX models, see [`mlx-router`](https://github.com/henrybravo/mlx-router).

```python
from mlx_lm import load, generate

model, tokenizer = load("henrybravo/NVIDIA-Nemotron-3-Nano-30B-A3B-4bit")

prompt = "hello"

if tokenizer.chat_template is not None:
    messages = [{"role": "user", "content": prompt}]
    prompt = tokenizer.apply_chat_template(
        messages, add_generation_prompt=True, return_dict=False,
    )

response = generate(model, tokenizer, prompt=prompt, verbose=True)
```