---
title: Nanbeige4.1-3B-heretic-4bit
canonical_url: "https://www.modelscope.cn/models/mlx-community/Nanbeige4.1-3B-heretic-4bit"
md_url: "https://www.modelscope.cn/models/mlx-community/Nanbeige4.1-3B-heretic-4bit.md"
repository: mlx-community/Nanbeige4.1-3B-heretic-4bit
last_updated: 2026-07-19
license: apache-2.0
pipeline_tag: text-generation
tasks:
  - text-generation
model_type:
  - llama
architectures:
  - LlamaForCausalLM
base_model:
  - Daizee/Nanbeige4.1-3B-heretic
base_model_relation: quantized
parameters: 614.8M
tensor_type:
  - BF16
  - U32
library_name:
  - mlx
  - safetensors
  - pytorch
frameworks:
  - pytorch
language:
  - en
  - zh
downloads: 202
stars: 0
tags:
  - llm
  - nanbeige
  - heretic
  - uncensored
  - decensored
  - abliterated
  - mlx
---

# Nanbeige4.1-3B-heretic-4bit

> Nanbeige4.1-3B-heretic-4bit - mlx-community 在 ModelScope 开源的模型。mlx-community/Nanbeige4.1-3B-heretic-4bit

mlx-community/Nanbeige4.1-3B-heretic-4bit 是 ModelScope 魔搭社区上的 614.8M 参数text-generation模型，采用 apache-2.0 许可，基于 Daizee/Nanbeige4.1-3B-heretic 构建。

- **Repository**: mlx-community/Nanbeige4.1-3B-heretic-4bit
- **License**: apache-2.0
- **Tasks**: text-generation
- **Parameters**: 614.8M
- **Base model**: Daizee/Nanbeige4.1-3B-heretic
- **Tags**: llm, nanbeige, heretic, uncensored, decensored, abliterated, mlx
- **Downloads**: 202
- **Stars**: 0
- **Last updated**: 2026-07-19

Source: https://www.modelscope.cn/models/mlx-community/Nanbeige4.1-3B-heretic-4bit

---

# mlx-community/Nanbeige4.1-3B-heretic-4bit

This model [mlx-community/Nanbeige4.1-3B-heretic-4bit](https://huggingface.co/mlx-community/Nanbeige4.1-3B-heretic-4bit) was
converted to MLX format from [Daizee/Nanbeige4.1-3B-heretic](https://huggingface.co/Daizee/Nanbeige4.1-3B-heretic)
using mlx-lm version **0.30.7**.

## Use with mlx

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

```python
from mlx_lm import load, generate

model, tokenizer = load("mlx-community/Nanbeige4.1-3B-heretic-4bit")

prompt = "hello"

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

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