z-lab / Llama-3.1-8B-Instruct-PARO

z-lab/Llama-3.1-8B-Instruct-PARO

Pairwise Rotation Quantization for Efficient Reasoning LLM Inference

ParoQuant is the state-of-the-art INT4 quantization for LLMs. It closes the accuracy gap with FP16 while running at near-AWQ speed. Supports NVIDIA GPUs (vLLM, Transformers) and Apple Silicon (MLX).

z-lab/Llama-3.1-8B-Instruct-PARO is a 4-bit meta-llama/Llama-3.1-8B-Instruct quantized with ParoQuant . Check out other ParoQuant models from the Hugging Face collection . Swap the model name in the commands below to try any of them.

Quick Start

Installation

# NVIDIA GPU
pip install "paroquant[vllm]"

# Apple Silicon
pip install "paroquant[mlx]"

Interactive Chat

python -m paroquant.cli.chat --model z-lab/Llama-3.1-8B-Instruct-PARO

OpenAI-Compatible API Server

python -m paroquant.cli.serve --model z-lab/Llama-3.1-8B-Instruct-PARO --port 8000

Docker (NVIDIA GPU)

# Interactive chat
docker run --pull=always --rm -it --gpus all --ipc=host \
  ghcr.io/z-lab/paroquant:chat --model z-lab/Llama-3.1-8B-Instruct-PARO

# API server (port 8000)
docker run --pull=always --rm -it --gpus all --ipc=host -p 8000:8000 \
  ghcr.io/z-lab/paroquant:serve --model z-lab/Llama-3.1-8B-Instruct-PARO

Citation

@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}
}