---
title: Sehyo-Qwen3.5-27B-NVFP4
canonical_url: "https://www.modelscope.cn/models/hf/Sehyo-Qwen3.5-27B-NVFP4"
md_url: "https://www.modelscope.cn/models/hf/Sehyo-Qwen3.5-27B-NVFP4.md"
repository: hf/Sehyo-Qwen3.5-27B-NVFP4
last_updated: 2026-03-02
model_type:
  - qwen3_5
architectures:
  - Qwen3_5ForConditionalGeneration
base_model:
  - Qwen/Qwen3.5-27B
base_model_relation: quantized
parameters: 19.6B
tensor_type:
  - F8_E4M3
  - BF16
  - F32
  - U8
library_name:
  - safetensors
  - pytorch
frameworks:
  - pytorch
downloads: 3477
stars: 3
tags:
  - quantized
  - nvfp4
  - compressed-tensors
  - llm-compressor
  - qwen3.5
---

# Sehyo-Qwen3.5-27B-NVFP4

> Sehyo-Qwen3.5-27B-NVFP4 - hf 在 ModelScope 开源的模型。This is a quantized version of Qwen/Qwen3.5-27B using the NVFP4 quantization scheme. Supports MTP and multi-modal.

hf/Sehyo-Qwen3.5-27B-NVFP4 是 ModelScope 魔搭社区上的 19.6B 参数机器学习模型，基于 Qwen/Qwen3.5-27B 构建。

- **Repository**: hf/Sehyo-Qwen3.5-27B-NVFP4
- **Parameters**: 19.6B
- **Base model**: Qwen/Qwen3.5-27B
- **Tags**: quantized, nvfp4, compressed-tensors, llm-compressor, qwen3.5
- **Downloads**: 3477
- **Stars**: 3
- **Last updated**: 2026-03-02

Source: https://www.modelscope.cn/models/hf/Sehyo-Qwen3.5-27B-NVFP4

---

# Qwen3.5-27B-NVFP4

This is a quantized version of [Qwen/Qwen3.5-27B](https://huggingface.co/Qwen/Qwen3.5-27B) using the **NVFP4** quantization scheme.
Supports MTP and multi-modal.

Please use nightly vLLM for support.

## Changelog
- **02/03/2026**: Initial upload.

## Calibration

- **Samples**: 128 (64 from each dataset) - This model is not a MoE model so high sample count not necessary.
- **Datasets**:
  - [HuggingFaceH4/ultrachat_200k](https://huggingface.co/datasets/HuggingFaceH4/ultrachat_200k) (`train_sft` split)
  - [nvidia/Nemotron-Post-Training-Dataset-v2](https://huggingface.co/datasets/nvidia/Nemotron-Post-Training-Dataset-v2) (`chat` split)
- **Max sequence length**: 4096

## Creation

This model was created using [VLLM's LLM Compressor](https://github.com/vllm-project/llm-compressor). The quantization ignores `lm_head`, `re:.*linear_attn.*`, and `re:model\.visual\..*` layers, preserving the vision encoder and linear attention modules at full precision. The model was loaded with `AutoModelForImageTextToText`.