---
title: llm-jp-4-8b-thinking-8bit
canonical_url: "https://www.modelscope.cn/models/mlx-community/llm-jp-4-8b-thinking-8bit"
md_url: "https://www.modelscope.cn/models/mlx-community/llm-jp-4-8b-thinking-8bit.md"
repository: mlx-community/llm-jp-4-8b-thinking-8bit
last_updated: 2026-04-12
license: apache-2.0
pipeline_tag: text-generation
tasks:
  - text-generation
model_type:
  - llama
architectures:
  - LlamaForCausalLM
base_model:
  - llm-jp/llm-jp-4-8b-thinking
base_model_relation: quantized
parameters: 2.4B
tensor_type:
  - U32
  - BF16
library_name:
  - mlx
  - safetensors
  - pytorch
frameworks:
  - pytorch
language:
  - en
  - ja
inference_backends:
  - "deploy_task text/emb"
  - "lmdeploy 0.9.1"
  - "lmdeploy_turbomind 0.9.1"
  - "sglang 0.5.2"
  - "vllm 0.9.2"
downloads: 8
stars: 0
tags:
  - mlx
---

# llm-jp-4-8b-thinking-8bit

> llm-jp-4-8b-thinking-8bit - mlx-community 在 ModelScope 开源的模型。mlx-community/llm-jp-4-8b-thinking-8bit

mlx-community/llm-jp-4-8b-thinking-8bit 是 ModelScope 魔搭社区上的 2.4B 参数text-generation模型，采用 apache-2.0 许可，基于 llm-jp/llm-jp-4-8b-thinking 构建，可用 deploy_task text/emb、lmdeploy 0.9.1、lmdeploy_turbomind 0.9.1 部署。

- **Repository**: mlx-community/llm-jp-4-8b-thinking-8bit
- **License**: apache-2.0
- **Tasks**: text-generation
- **Parameters**: 2.4B
- **Base model**: llm-jp/llm-jp-4-8b-thinking
- **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**: 8
- **Stars**: 0
- **Last updated**: 2026-04-12

Source: https://www.modelscope.cn/models/mlx-community/llm-jp-4-8b-thinking-8bit

---

# mlx-community/llm-jp-4-8b-thinking-8bit

This model [mlx-community/llm-jp-4-8b-thinking-8bit](https://huggingface.co/mlx-community/llm-jp-4-8b-thinking-8bit) was
converted to MLX format from [llm-jp/llm-jp-4-8b-thinking](https://huggingface.co/llm-jp/llm-jp-4-8b-thinking)
using mlx-lm version **0.31.1**.

## Use with mlx (not support harmony format)

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

```python
from mlx_lm import load, generate

model, tokenizer = load("mlx-community/llm-jp-4-8b-thinking-8bit")

prompt = "こんにちは"

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)
```

## Use with vllm-mlx (support harmony format)

```bash
pip install git+https://github.com/waybarrios/vllm-mlx.git
vllm-mlx serve mlx-community/llm-jp-4-8b-thinking-8bit --reasoning-parser gpt_oss --port 8000
vllm-mlx-chat --server-url http://localhost:8000
```
or,

```python
from openai import OpenAI

# Without API key (local development)
client = OpenAI(base_url="http://localhost:8000/v1", api_key="not-needed")

response = client.chat.completions.create(
    model="default",
    messages=[{"role": "user", "content": "こんにちは"}],
)
print(response.choices[0].message.content)
```