You can deploy this model and start earning money today!

HarimxChoi/
WarpQuant-Qwen3.5-4B-R16E4V4

Name: WarpQuant-Qwen3.5-4B-R16E4V4
Availability: InStock

Views

No views yet

WarpQuant Qwen3.5-4B R16E4V4

Qwen3.5-4B quantized with signed Hadamard rotation, block-GPTQ, and Output-Fisher weak-column recovery. The projection layers use a 3.5-bpw INT3 base, selected columns are restored in BF16, and the token embedding and vision weights use group-128 INT4.

Technical report · GitHub · Text-only model

Format

Component	Format
Text projections	signed Hadamard + block-GPTQ INT3
Weak columns	Output-Fisher selected BF16 recovery
Token embedding	group-128 INT4
Vision weights	group-128 INT4
Full-model packed-equivalent rate	3.6912 bpw

The repository stores the quantized values in BF16-compatible safetensors for direct Transformers loading. The bpw value is the analytical packed payload including codes, scales, recovery values, and column indices.

Evaluation

Scope	Metric	Result
Text backbone	WikiText-2 PPL ↓	9.2494
Text backbone	ARC-Challenge, 299 ↑	46.15
Text backbone	MMLU, 13,943 ↑	38.13
Full VLM	ChartQA-128 exact ↑	44.53
Full VLM	ChartQA-128 relaxed ↑	53.91

Usage

Qwen3.5 currently requires the latest Transformers main branch:

pip install "transformers @ git+https://github.com/huggingface/transformers.git@main"

python

import torch
from transformers import AutoModelForImageTextToText, AutoProcessor

model_id = "HarimxChoi/WarpQuant-Qwen3.5-4B-R16E4V4"
processor = AutoProcessor.from_pretrained(model_id)
model = AutoModelForImageTextToText.from_pretrained(
    model_id,
    torch_dtype=torch.bfloat16,
    device_map="auto",
)

Citation

bibtex

@misc{choi2026warpquant,
  author = {Harim Choi},
  title = {WarpQuant: Dual-Domain LLM Quantization via Hadamard Rotation and Output-Fisher Sensitivity},
  year = {2026},
  url = {https://harimxchoi.github.io/projects/warpquant/}
}