z-lab/gemma-4-E4B-it-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). For more information, see https://github.com/z-lab/paroquant.

z-lab/gemma-4-E4B-it-PARO is a 4-bit google/gemma-4-E4B-it quantized with ParoQuant. Check out other ParoQuant models from the Hugging Face collection.

Quick Start

Installation

# NVIDIA GPU (CUDA 12.9)
pip install "paroquant[vllm]"

# NVIDIA GPU (CUDA 13.0)
pip install "paroquant[vllm]" "vllm==0.19.1" \
  --extra-index-url https://wheels.vllm.ai/0.19.1/cu130 \
  --extra-index-url https://download.pytorch.org/whl/cu130

# Apple Silicon
pip install "paroquant[mlx]"

Interactive Chat

python -m paroquant.cli.chat --model z-lab/gemma-4-E4B-it-PARO

OpenAI-Compatible API Server

For vLLM, you can directly use vllm serve to serve ParoQuant models:

vllm serve z-lab/gemma-4-E4B-it-PARO --port 8000

For other frameworks:

python -m paroquant.cli.serve --model z-lab/gemma-4-E4B-it-PARO --port 8000

For MLX, add --vlm if you wish to load the VLM components and use the model's multimodal features. For vLLM, VLM components are loaded by default and can be skipped with the server argument --language-model-only.

The visual components in this checkpoint is stored in original precision, and only the language components are quantized to 4 bits; as a result, the model size is larger than a fully-quantized model. Avoid loading the VLM components if you are not using the multimodal features for the best efficiency.

Docker (NVIDIA GPU)

The following commands map the local cache directory to the container in order to persist kernel cache across runs. Remove -v ... to disable this behavior.

# Interactive chat
docker run --pull=always --rm -it --gpus all --ipc=host \
  -v $HOME/.cache/paroquant:/root/.cache/paroquant \
  ghcr.io/z-lab/paroquant:chat --model z-lab/gemma-4-E4B-it-PARO

# API server (port 8000)
docker run --pull=always --rm -it --gpus all --ipc=host -p 8000:8000 \
  -v $HOME/.cache/paroquant:/root/.cache/paroquant \
  ghcr.io/z-lab/paroquant:serve --model z-lab/gemma-4-E4B-it-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}
}