---
title: Qwen3.5-2B-MLX-4bit
canonical_url: "https://www.modelscope.cn/models/mlx-community/Qwen3.5-2B-MLX-4bit"
md_url: "https://www.modelscope.cn/models/mlx-community/Qwen3.5-2B-MLX-4bit.md"
repository: mlx-community/Qwen3.5-2B-MLX-4bit
last_updated: 2026-03-02
license: apache-2.0
model_type:
  - qwen3_5
architectures:
  - Qwen3_5ForConditionalGeneration
base_model:
  - Qwen/Qwen3.5-2B
base_model_relation: quantized
parameters: 625.9M
tensor_type:
  - BF16
  - U32
  - F32
library_name:
  - mlx
  - safetensors
  - pytorch
frameworks:
  - pytorch
downloads: 645
stars: 1
tags:
  - mlx
  - qwen3.5
  - vision-language-model
  - quantized
  - 4bit
---

# Qwen3.5-2B-MLX-4bit

> Qwen3.5-2B-MLX-4bit - mlx-community 在 ModelScope 开源的模型。This is a 4-bit quantized MLX version of Qwen/Qwen3.5-2B for Apple Silicon.

mlx-community/Qwen3.5-2B-MLX-4bit 是 ModelScope 魔搭社区上的 625.9M 参数机器学习模型，采用 apache-2.0 许可，基于 Qwen/Qwen3.5-2B 构建。

- **Repository**: mlx-community/Qwen3.5-2B-MLX-4bit
- **License**: apache-2.0
- **Parameters**: 625.9M
- **Base model**: Qwen/Qwen3.5-2B
- **Tags**: mlx, qwen3.5, vision-language-model, quantized, 4bit
- **Downloads**: 645
- **Stars**: 1
- **Last updated**: 2026-03-02

Source: https://www.modelscope.cn/models/mlx-community/Qwen3.5-2B-MLX-4bit

---

# Qwen3.5-2B-MLX-4bit

This is a 4-bit quantized MLX version of [Qwen/Qwen3.5-2B](https://huggingface.co/Qwen/Qwen3.5-2B) for Apple Silicon.

## Model Details

- **Original Model:** [Qwen/Qwen3.5-2B](https://huggingface.co/Qwen/Qwen3.5-2B)
- **Quantization:** 4-bit (6.225 bits per weight)
- **Group Size:** 64
- **Format:** MLX SafeTensors
- **Framework:** [mlx-vlm](https://github.com/Blaizzy/mlx-vlm)
- **Disk Size:** ~1.6G

## Conversion Details

This model was converted using `mlx-vlm` from the [`pc/fix-qwen35-predicate`](https://github.com/Blaizzy/mlx-vlm/tree/pc/fix-qwen35-predicate) branch, which includes fixes for Qwen3.5 model support (proper handling of MoE gate layers, `shared_expert_gate`, and `A_log` casting).

**Conversion command:**
```bash
python3 -m mlx_vlm convert \
  --hf-path "Qwen/Qwen3.5-2B" \
  --mlx-path "./Qwen3.5-2B-MLX-4bit" \
  -q --q-bits 4 --q-group-size 64
```

## Important Note

A better, more optimized conversion may be available from **@Prince** ([@Blaizzy](https://huggingface.co/Blaizzy)) in the MLX VLM community. Check the [mlx-community](https://huggingface.co/mlx-community) organization for updated versions as official Qwen3.5 support is merged into the main `mlx-vlm` branch.

## Related Models

- **bf16 (full precision):** [mlx-community/Qwen3.5-2B-MLX-bf16](https://huggingface.co/mlx-community/Qwen3.5-2B-MLX-bf16)
- **8-bit quantized:** [mlx-community/Qwen3.5-2B-MLX-8bit](https://huggingface.co/mlx-community/Qwen3.5-2B-MLX-8bit)
- **Original:** [Qwen/Qwen3.5-2B](https://huggingface.co/Qwen/Qwen3.5-2B)

## Usage

```python
from mlx_vlm import load, generate

model, processor = load("mlx-community/Qwen3.5-2B-MLX-4bit")

output = generate(
    model,
    processor,
    prompt="Describe this image.",
    image="path/to/image.jpg",
    max_tokens=512
)
print(output)
```

**CLI:**
```bash
python3 -m mlx_vlm.generate \
  --model mlx-community/Qwen3.5-2B-MLX-4bit \
  --image path/to/image.jpg \
  --prompt "Describe this image."
```

## License

This model inherits the [Apache 2.0 license](https://huggingface.co/Qwen/Qwen3.5-2B) from the original Qwen model.