---
title: FuseO1-QwQ-DeepSeekR1-LightR1-32B-3bit
canonical_url: "https://www.modelscope.cn/models/mlx-community/FuseO1-QwQ-DeepSeekR1-LightR1-32B-3bit"
md_url: "https://www.modelscope.cn/models/mlx-community/FuseO1-QwQ-DeepSeekR1-LightR1-32B-3bit.md"
repository: mlx-community/FuseO1-QwQ-DeepSeekR1-LightR1-32B-3bit
last_updated: 2025-03-10
license: apache-2.0
model_type:
  - qwen2
architectures:
  - Qwen2ForCausalLM
base_model:
  - FuseAI/FuseO1-QwQ-DeepSeekR1-LightR1-32B
base_model_relation: quantized
parameters: 4.1B
tensor_type:
  - F16
  - U32
library_name:
  - mlx
  - safetensors
  - pytorch
frameworks:
  - pytorch
inference_backends:
  - "deploy_task text/emb"
  - "lmdeploy 0.9.1"
  - "lmdeploy_turbomind 0.9.1"
  - "sglang 0.5.2"
  - "vllm 0.9.2"
downloads: 76
stars: 0
tags:
  - mlx
---

# FuseO1-QwQ-DeepSeekR1-LightR1-32B-3bit

> FuseO1-QwQ-DeepSeekR1-LightR1-32B-3bit - mlx-community 在 ModelScope 开源的模型。3-bit mlx works, and better than I expected.

mlx-community/FuseO1-QwQ-DeepSeekR1-LightR1-32B-3bit 是 ModelScope 魔搭社区上的 4.1B 参数机器学习模型，采用 apache-2.0 许可，基于 FuseAI/FuseO1-QwQ-DeepSeekR1-LightR1-32B 构建，可用 deploy_task text/emb、lmdeploy 0.9.1、lmdeploy_turbomind 0.9.1 部署。

- **Repository**: mlx-community/FuseO1-QwQ-DeepSeekR1-LightR1-32B-3bit
- **License**: apache-2.0
- **Parameters**: 4.1B
- **Base model**: FuseAI/FuseO1-QwQ-DeepSeekR1-LightR1-32B
- **Inference backends**: deploy_task text/emb, lmdeploy 0.9.1, lmdeploy_turbomind 0.9.1, sglang 0.5.2, vllm 0.9.2
- **Tags**: mlx
- **Downloads**: 76
- **Stars**: 0
- **Last updated**: 2025-03-10

Source: https://www.modelscope.cn/models/mlx-community/FuseO1-QwQ-DeepSeekR1-LightR1-32B-3bit

---

3-bit mlx works, and better than I expected.

# bobig/FuseO1-QwQ-DeepSeekR1-LightR1-32B-3bit

The Model [bobig/FuseO1-QwQ-DeepSeekR1-LightR1-32B-3bit](https://huggingface.co/bobig/FuseO1-QwQ-DeepSeekR1-LightR1-32B-3bit) was converted to MLX format from [FuseAI/FuseO1-QwQ-DeepSeekR1-LightR1-32B](https://huggingface.co/FuseAI/FuseO1-QwQ-DeepSeekR1-LightR1-32B) using mlx-lm version **0.21.5**.

## Use with mlx

```bash
pip install mlx-lm
```

```python
from mlx_lm import load, generate

model, tokenizer = load("bobig/FuseO1-QwQ-DeepSeekR1-LightR1-32B-3bit")

prompt="hello"

if hasattr(tokenizer, "apply_chat_template") and tokenizer.chat_template is not None:
    messages = [{"role": "user", "content": prompt}]
    prompt = tokenizer.apply_chat_template(
        messages, tokenize=False, add_generation_prompt=True
    )

response = generate(model, tokenizer, prompt=prompt, verbose=True)
```