---
title: EXAONE-4.0.1-32B-MLX-Q4
canonical_url: "https://www.modelscope.cn/models/mlx-community/EXAONE-4.0.1-32B-MLX-Q4"
md_url: "https://www.modelscope.cn/models/mlx-community/EXAONE-4.0.1-32B-MLX-Q4.md"
repository: mlx-community/EXAONE-4.0.1-32B-MLX-Q4
last_updated: 2026-01-20
license: other
pipeline_tag: text-generation
tasks:
  - text-generation
  - text2text-generation
model_type:
  - exaone4
architectures:
  - Exaone4ForCausalLM
base_model:
  - LGAI-EXAONE/EXAONE-4.0-32B
base_model_relation: quantized
parameters: 557.3M
tensor_type:
  - BF16
  - U32
library_name:
  - safetensors
  - pytorch
frameworks:
  - pytorch
language:
  - en
  - ko
  - es
downloads: 19
stars: 0
---

# EXAONE-4.0.1-32B-MLX-Q4

> EXAONE-4.0.1-32B-MLX-Q4 - mlx-community 在 ModelScope 开源的模型。[!IMPORTANT] Original Model Link : https://huggingface.co/LGAI-EXAONE/EXAONE-4.0.1-32B

mlx-community/EXAONE-4.0.1-32B-MLX-Q4 是 ModelScope 魔搭社区上的 557.3M 参数text-generation、text2text-generation模型，采用 other 许可，基于 LGAI-EXAONE/EXAONE-4.0-32B 构建。

- **Repository**: mlx-community/EXAONE-4.0.1-32B-MLX-Q4
- **License**: other
- **Tasks**: text-generation, text2text-generation
- **Parameters**: 557.3M
- **Base model**: LGAI-EXAONE/EXAONE-4.0-32B
- **Downloads**: 19
- **Stars**: 0
- **Last updated**: 2026-01-20

Source: https://www.modelscope.cn/models/mlx-community/EXAONE-4.0.1-32B-MLX-Q4

---

> [!IMPORTANT]
> Original Model Link : [https://huggingface.co/LGAI-EXAONE/EXAONE-4.0.1-32B](https://huggingface.co/LGAI-EXAONE/EXAONE-4.0.1-32B)
>

```
name: EXAONE-4.0.1-32B-MLX-Q4
base_model: LGAI-EXAONE/EXAONE-4.0-32B
license: other
pipeline_tag: text-generation
tasks :
 - text-generation
 - text2text-generation
language :
 - en
 - ko
 - es
license_link: LICENSE
get_started_code: uvx --from mlx-lm mlx_lm.generate --model  --model mlx-community/EXAONE-4.0.1-32B-MLX-Q4 --prompt 'Test Prompt' --prompt 'Test prompt'
```


# EXAONE-4.0.1-32B-MLX-Q4
EXAONE-4.0.1-32B-MLX-Q4 is a hybrid thinking and non-thinking model with tool use from LGAI. A multilingual model, it supports English, Korean and Spanish, and works best at `temperature<0.6` in all modes and `top_p=0.95` for thinking. `presence_penalty=1.5` is recommended in the case of model degeneration, and `temperature=0.1` to force answers to a single language.

> [!WARNING]
> MLX is a framework for METAL graphics supported by Apple computers with ARM M-series processors (M1/M2/M3/M4)

> [!NOTE]
> Generation using uv https://docs.astral.sh/uv/**:
> ```
> uvx --from mlx-lm mlx_lm.generate --model  --model mlx-community/EXAONE-4.0.1-32B-MLX-Q4 --prompt 'Test Prompt'
>```

> [!NOTE]
> Generation using pip:
> ```
> pipx --from mlx-lm mlx_lm.generate --model  --model mlx-community/EXAONE-4.0.1-32B-MLX-Q4 --prompt 'Test Prompt'
> ```