README.md

---
license: other
license_name: nvidia-license
license_link: https://huggingface.co/nvidia/Alpamayo-1.5-10B/blob/main/LICENSE
pipeline_tag: robotics
language:
- en
inference: false
base_model:
- z-lab/Alpamayo-R1-10B
base_model_relation: quantized
tags:
- flashdrive
- paroquant
- quantization
- w4a8
- autonomous-driving
- vision-language-action
---

# Alpamayo 1 (R1) — W4A8 (ParoQuant)

**Flash Vision-Language-Action Inference for Autonomous Driving**

[![Paper](https://img.shields.io/badge/arXiv-2511.10645-b31b1b.svg)](https://arxiv.org/abs/2511.10645)
[![GitHub](https://img.shields.io/badge/GitHub-FlashDrive-181717?logo=github)](https://github.com/z-lab/flashdrive)
[![Blog](https://img.shields.io/badge/Blog-FlashDrive-blue)](https://z-lab.ai/projects/flashdrive/)
[![Models](https://img.shields.io/badge/%F0%9F%A4%97-Models-yellow)](https://huggingface.co/collections/z-lab/flashdrive)

W4A8 [ParoQuant](https://github.com/z-lab/paroquant) weights for the language model of [z-lab/Alpamayo-R1-10B](https://huggingface.co/z-lab/Alpamayo-R1-10B), used by [FlashDrive](https://github.com/z-lab/flashdrive) to accelerate [Alpamayo 1 (R1)](https://huggingface.co/nvidia/Alpamayo-R1-10B).

ParoQuant (ICLR 2026) is a state-of-the-art INT4 quantizer: learned pairwise rotations suppress activation outliers, closing the accuracy gap with FP16 at near-AWQ speed. Here it quantizes the VLM language model to INT4 weights and INT8 activations (served through vLLM's Marlin kernels); the action expert stays bf16.

> [!NOTE]
> **Not a standalone model.** FlashDrive loads the [base checkpoint](https://huggingface.co/z-lab/Alpamayo-R1-10B) and fills these quantized weights automatically — you do not load this repository directly.

## Usage

```python
import flashdrive

# from_pretrained fetches this -PARO checkpoint automatically
model = flashdrive.from_pretrained("z-lab/Alpamayo-R1-10B")
```

See the [base model card](https://huggingface.co/z-lab/Alpamayo-R1-10B) and the [FlashDrive repository](https://github.com/z-lab/flashdrive) for the full pipeline.

## License

This checkpoint is derived from NVIDIA's Alpamayo weights and is governed by the [NVIDIA License](https://huggingface.co/nvidia/Alpamayo-1.5-10B/blob/main/LICENSE), which permits **non-commercial use only** and extends to derivative works. The [FlashDrive](https://github.com/z-lab/flashdrive) inference code is separately released under the [MIT License](https://github.com/z-lab/flashdrive/blob/main/LICENSE).

## Citation

```bibtex
@inproceedings{liang2026paroquant,
  title     = {{ParoQuant: Pairwise Rotation Quantization for Efficient Reasoning LLM Inference}},
  author    = {Liang, Yesheng and Chen, Haisheng and Zhang, Zihan and Han, Song and Liu, Zhijian},
  booktitle = {International Conference on Learning Representations (ICLR)},
  year      = {2026}
}
```

```bibtex
@article{li2026flashdrive,
  title  = {{FlashDrive: Flash Vision-Language-Action Inference for Autonomous Driving}},
  author = {Li, Zekai and Liang, Yihao and Zhang, Hongfei and Chen, Jian and Liang, Yesheng and Liu, Zhijian},
  year   = {2026}
}
```