---
title: CAT-Thinking-8B-MLX-6bit
canonical_url: "https://www.modelscope.cn/models/mlx-community/CAT-Thinking-8B-MLX-6bit"
md_url: "https://www.modelscope.cn/models/mlx-community/CAT-Thinking-8B-MLX-6bit.md"
repository: mlx-community/CAT-Thinking-8B-MLX-6bit
last_updated: 2026-06-03
license: apache-2.0
pipeline_tag: text-generation
tasks:
  - text-generation
model_type:
  - qwen3
architectures:
  - Qwen3ForCausalLM
base_model:
  - cyberagent/CAT-Thinking-8B
base_model_relation: quantized
parameters: 1.8B
tensor_type:
  - U32
  - BF16
library_name:
  - mlx
  - safetensors
  - pytorch
frameworks:
  - pytorch
language:
  - en
  - ja
downloads: 12
stars: 0
tags:
  - mlx
---

# CAT-Thinking-8B-MLX-6bit

> CAT-Thinking-8B-MLX-6bit - mlx-community 在 ModelScope 开源的模型。How to use git clone https://github.com/ml-explore/mlx-lm (mlx-lm==0.31.3)

mlx-community/CAT-Thinking-8B-MLX-6bit 是 ModelScope 魔搭社区上的 1.8B 参数text-generation模型，采用 apache-2.0 许可，基于 cyberagent/CAT-Thinking-8B 构建。

- **Repository**: mlx-community/CAT-Thinking-8B-MLX-6bit
- **License**: apache-2.0
- **Tasks**: text-generation
- **Parameters**: 1.8B
- **Base model**: cyberagent/CAT-Thinking-8B
- **Tags**: mlx
- **Downloads**: 12
- **Stars**: 0
- **Last updated**: 2026-06-03

Source: https://www.modelscope.cn/models/mlx-community/CAT-Thinking-8B-MLX-6bit

---

How to use<br>
git clone https://github.com/ml-explore/mlx-lm 
(mlx-lm==0.31.3)

> mlx_lm.convert --hf-path cyberagent/CAT-Thinking-8B --mlx-path ./CAT-Thinking-8B-MLX-6bit  -q --q-bits 6 --trust-remote-code

> mlx_lm.generate --model ./CAT-Thinking-8B-MLX-6bit --verbose True  --prompt "about you"  --max-tokens 1000

==========<br>
Prompt: 10 tokens, 90.942 tokens-per-sec<br>
Generation: 949 tokens, 84.891 tokens-per-sec<br>
Peak memory: 6.865 GB