---
title: Klear-46B-A2.5B-Instruct-3bit
canonical_url: "https://www.modelscope.cn/models/mlx-community/Klear-46B-A2.5B-Instruct-3bit"
md_url: "https://www.modelscope.cn/models/mlx-community/Klear-46B-A2.5B-Instruct-3bit.md"
repository: mlx-community/Klear-46B-A2.5B-Instruct-3bit
last_updated: 2025-09-08
license: apache-2.0
pipeline_tag: text-generation
tasks:
  - text-generation
model_type:
  - Klear
architectures:
  - KlearMoeForCausalLM
base_model:
  - Kwai-Klear/Klear-46B-A2.5B-Instruct
base_model_relation: quantized
parameters: 5.8B
tensor_type:
  - BF16
  - U32
  - F32
library_name:
  - mlx
  - safetensors
  - pytorch
frameworks:
  - pytorch
language:
  - zh
  - en
downloads: 78
stars: 0
tags:
  - mlx
---

# Klear-46B-A2.5B-Instruct-3bit

> Klear-46B-A2.5B-Instruct-3bit - mlx-community 在 ModelScope 开源的模型。mlx-community/Klear-46B-A2.5B-Instruct-3bit

mlx-community/Klear-46B-A2.5B-Instruct-3bit 是 ModelScope 魔搭社区上的 5.8B 参数text-generation模型，采用 apache-2.0 许可，基于 Kwai-Klear/Klear-46B-A2.5B-Instruct 构建。

- **Repository**: mlx-community/Klear-46B-A2.5B-Instruct-3bit
- **License**: apache-2.0
- **Tasks**: text-generation
- **Parameters**: 5.8B
- **Base model**: Kwai-Klear/Klear-46B-A2.5B-Instruct
- **Tags**: mlx
- **Downloads**: 78
- **Stars**: 0
- **Last updated**: 2025-09-08

Source: https://www.modelscope.cn/models/mlx-community/Klear-46B-A2.5B-Instruct-3bit

---

# mlx-community/Klear-46B-A2.5B-Instruct-3bit

This model [mlx-community/Klear-46B-A2.5B-Instruct-3bit](https://huggingface.co/mlx-community/Klear-46B-A2.5B-Instruct-3bit) was
converted to MLX format from [Kwai-Klear/Klear-46B-A2.5B-Instruct](https://huggingface.co/Kwai-Klear/Klear-46B-A2.5B-Instruct)
using mlx-lm version **0.27.1**.

## Use with mlx

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

```python
from mlx_lm import load, generate

model, tokenizer = load("mlx-community/Klear-46B-A2.5B-Instruct-3bit")

prompt = "hello"

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

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