Hugging Face's logo

K-intelligence
/
Midm-2.0-Mini-Instruct

Text Generation
Transformers
Safetensors
English
Korean
llama
KT
K-intelligence
Mi:dm
conversational
text-generation-inference
Model card Files
xet
Community
Midm-2.0-Mini-Instruct / README.md
Midm-LLM's picture
Midm-LLM
Initial upload
a4e305d verified
preview code
|
raw
history blame
16 kB
 ---
license: mit
language:
- en
- ko
tags:
- KT
- K-intelligence
- Mi:dm
pipeline_tag: text-generation
library_name: transformers
---
<p align="center">
<br>
<span style="font-size: 60px; font-weight: bold;">Mi:dm 2.0-Mini</span>
</br>
</p>
<p align="center">
🤗 <a href="">Mi:dm 2.0 Models</a> |
📜 Mi:dm 2.0 Technical Report* |
📕 Mi:dm 2.0 Technical Blog*
</p>
<p align="center"><sub>*To be released soon</sub></p>
<br>
## News 📢
- 🔜 _(Coming Soon!) GGUF format model files will be available soon for easier local deployment._
- ⚡️`2025/07/04`: Released Mi:dm 2.0 Model collection on Hugging Face🤗.
<br>
<br>
# Table of Contents
- ___Overview___
- [Mi:dm 2.0](#midm-20)
- [Quickstart](#quickstart)
- [Evaluation](#evaluation)
- ___Usage___
- [Run on Friendly.AI](#run-on-friendliai)
- [Run on Your Local Machine](#run-on-your-local-machine)
- [Deployment](#deployment)
- [Tutorials](#tutorials)
- ___More Information___
- [Limitation](#limitation)
- [License](#license)
- [Contact](#contact)
<br>
<br>
# Overview
### Mi:dm 2.0
Mi:dm 2.0 is a __"Korean-centric AI"__ model developed with KT's proprietary technology. __"Korean-centric AI"__ refers to a model that thoroughly internalizes the unique values, cognitive frameworks, and commonsense reasoning intrinsic to Korean society. It is not simply about processing and responding in Korean; it is about the profound understanding that reflects and respects the socio-cultural fabric of Korean norms and values.
The newly introduced Mi:dm 2.0 model comes in two versions:
* **Mi:dm 2.0-Mini** is a 2.3B parameter Dense small model, designed for seamless use in environments such as on-device settings and low-end GPUs. It was created by pruning and distilling the Base model.
* **Mi:dm 2.0-Base** has 11.5B parameters and was designed to balance model size and performance by expanding an 8B scale model using the DuS (Depth-up Scaling) method. It's a practical model that can be applied to various real-world services, considering both performance and versatility.
> [!Note]
> Neither the pre-training nor the post-training data includes KT users' data.
<br>
### Quickstart
Here is the code snippet to run conversational inference with the model:
```python
import torch
from transformers import AutoModelForCausalLM, AutoTokenizer, GenerationConfig
model_name = "K-intelligence/Midm-2.0-Mini-Instruct"
model = AutoModelForCausalLM.from_pretrained(
model_name,
torch_dtype=torch.bfloat16,
trust_remote_code=True,
device_map="auto"
)
tokenizer = AutoTokenizer.from_pretrained(model_name)
generation_config = GenerationConfig.from_pretrained(model_name)
prompt = "KT에 대해 소개해줘"
# message for inference
messages = [
{"role": "system",
"content": "Mi:dm(믿:음)은 KT에서 개발한 AI 기반 어시스턴트이다."},
{"role": "user", "content": prompt}
]
input_ids = tokenizer.apply_chat_template(
messages,
tokenize=True,
add_generation_prompt=True,
return_tensors="pt"
)
output = model.generate(
input_ids.to("cuda"),
generation_config=generation_config,
eos_token_id=tokenizer.eos_token_id,
max_new_tokens=128,
do_sample=False,
)
print(tokenizer.decode(output[0]))
```
> [!NOTE]
> The `transformers` library should be version `4.45.0` or higher.
<br>
<br>
# Evaluation
#### English
<table>
<thead>
<tr>
<th colspan="2"><b>Benchmark</b></th>
<th>Exaone-3.5-2.4B-inst</th>
<th>Qwen3-4B</th>
<th>Mi:dm 2.0-Mini-inst</th>
<th>Exaone-3.5-7.8B-inst</th>
<th>Qwen3-14B</th>
<th>Llama-3.1-8B-inst</th>
<th>Mi:dm 2.0-Base-inst</th>
</tr>
</thead>
<tbody>
<tr>
<td rowspan="1"><b>Instruction Following</b></td>
<td><b>IFEval</b></td>
<td align="center">81.1</td>
<td align="center">79.7</td>
<td align="center">73.6</td>
<td align="center">83.6</td>
<td align="center">83.9</td>
<td align="center">79.9</td>
<td align="center"><b>84.0</b></td>
</tr>
<tr>
<td rowspan="4"><b>Reasoning</b></td>
<td><b>BBH</b></td>
<td align="center">46.4</td>
<td align="center">79.0</td>
<td align="center">44.5</td>
<td align="center">50.1</td>
<td align="center">83.4</td>
<td align="center">60.3</td>
<td align="center"><b>77.7</b></td>
</tr>
<tr>
<td><b>GPQA</b></td>
<td align="center">28.1</td>
<td align="center">39.8</td>
<td align="center">26.6</td>
<td align="center">33.1</td>
<td align="center">49.8</td>
<td align="center">21.6</td>
<td align="center"><b>33.5</b></td>
</tr>
<tr>
<td><b>MuSR</b></td>
<td align="center">49.7</td>
<td align="center">58.5</td>
<td align="center">51.7</td>
<td align="center">51.2</td>
<td align="center">57.7</td>
<td align="center">50.3</td>
<td align="center"><b>51.9</b></td>
</tr>
<tr>
<td><b>Avg.</b></td>
<td align="center">41.4</td>
<td align="center">59.1</td>
<td align="center">40.9</td>
<td align="center">44.8</td>
<td align="center">63.6</td>
<td align="center">44.1</td>
<td align="center"><b>54.4</b></td>
</tr>
<tr>
<td rowspan="2"><b>Mathematics</b></td>
<td><b>GSM8K</b></td>
<td align="center">82.5</td>
<td align="center">90.4</td>
<td align="center">83.1</td>
<td align="center">81.1</td>
<td align="center">88.0</td>
<td align="center">81.2</td>
<td align="center"><b>91.6</b></td>
</tr>
<tr>
<td><b>MBPP+</b></td>
<td align="center">59.8</td>
<td align="center">62.4</td>
<td align="center">60.9</td>
<td align="center">79.4</td>
<td align="center">73.4</td>
<td align="center">81.8</td>
<td align="center"><b>77.5</b></td>
</tr>
<tr>
<td rowspan="3"><b>General Knowledge</b></td>
<td><b>MMLU-pro</b></td>
<td align="center">-</td>
<td align="center">-</td>
<td align="center">-</td>
<td align="center">40.7</td>
<td align="center">70.5</td>
<td align="center">47.6</td>
<td align="center"><b>53.3</b></td>
</tr>
<tr>
<td><b>MMLU</b></td>
<td align="center">59.5</td>
<td align="center">73.3</td>
<td align="center">56.5</td>
<td align="center">69.0</td>
<td align="center">82.7</td>
<td align="center">70.7</td>
<td align="center"><b>73.7</b></td>
</tr>
<tr>
<td><b>Avg.</b></td>
<td align="center">59.5</td>
<td align="center">73.3</td>
<td align="center">56.5</td>
<td align="center">54.8</td>
<td align="center"><b>76.6</b></td>
<td align="center">59.2</td>
<td align="center">63.5</td>
</tr>
</tbody>
</table>
#### Korean
<table>
<thead>
<tr>
<th colspan="2"><b>Benchmark</b></th>
<th>Exaone-3.5-2.4B-inst</th>
<th>Qwen3-4B</th>
<th>Mi:dm 2.0-Mini-inst</th>
<th>Exaone-3.5-7.8B-inst</th>
<th>Qwen3-14B</th>
<th>Llama-3.1-8B-inst</th>
<th>Mi:dm 2.0-Base-inst</th>
</tr>
</thead>
<tbody>
<!-- Comprehension -->
<tr>
<td rowspan="5"><b>Comprehension</b></td>
<td><b>K-Prag*</b></td>
<td align="center">68.7</td>
<td align="center">73.9</td>
<td align="center">69.5</td>
<td align="center">73.5</td>
<td align="center"><b>86.7</b></td>
<td align="center">59.9</td>
<td align="center">86.5</td>
</tr>
<tr>
<td><b>K-Refer-Hard*</b></td>
<td align="center">58.5</td>
<td align="center">56.7</td>
<td align="center">55.4</td>
<td align="center">61.9</td>
<td align="center"><b>74.0</b></td>
<td align="center">48.6</td>
<td align="center">70.8</td>
</tr>
<tr>
<td><b>Ko-Best</b></td>
<td align="center">87.2</td>
<td align="center">91.5</td>
<td align="center">80.5</td>
<td align="center">92.0</td>
<td align="center">93.9</td>
<td align="center">77.4</td>
<td align="center"><b>95.2</b></td>
</tr>
<tr>
<td><b>Ko-Sovereign*</b></td>
<td align="center">38.0</td>
<td align="center">43.5</td>
<td align="center">42.5</td>
<td align="center">44.0</td>
<td align="center">52.0</td>
<td align="center">31.5</td>
<td align="center"><b>53.0</b></td>
</tr>
<tr>
<td><b>Avg.</b></td>
<td align="center">62.5</td>
<td align="center">66.6</td>
<td align="center">61.9</td>
<td align="center">67.2</td>
<td align="center"><b>76.8</b></td>
<td align="center">51.5</td>
<td align="center">76.1</td>
</tr>
<tr>
<td rowspan="5"><b>Reasoning</b></td>
<td><b>Ko-Winogrande</b></td>
<td align="center">60.3</td>
<td align="center"><b>67.5</b></td>
<td align="center">61.7</td>
<td align="center">64.6</td>
<td align="center">77.2</td>
<td align="center">40.1</td>
<td align="center">75.1</td>
</tr>
<tr>
<td><b>Ko-Best</b></td>
<td align="center">64.1</td>
<td align="center"><b>69.2</b></td>
<td align="center">64.5</td>
<td align="center">60.3</td>
<td align="center">75.4</td>
<td align="center">26.0</td>
<td align="center">73.0</td>
</tr>
<tr>
<td><b>LogicKor*</b></td>
<td align="center"><b>7.4</b></td>
<td align="center">5.6</td>
<td align="center">7.7</td>
<td align="center">8.6</td>
<td align="center">6.4</td>
<td align="center">2.4</td>
<td align="center">8.6</td>
</tr>
<tr>
<td><b>HRM8K*</b></td>
<td align="center">38.5</td>
<td align="center"><b>56.7</b></td>
<td align="center">39.9</td>
<td align="center">49.7</td>
<td align="center">64.5</td>
<td align="center">30.9</td>
<td align="center">52.9</td>
</tr>
<tr>
<td><b>Avg.</b></td>
<td align="center">36.7</td>
<td align="center"><b>43.8</b></td>
<td align="center">37.4</td>
<td align="center">39.5</td>
<td align="center">48.8</td>
<td align="center">19.8</td>
<td align="center">44.8</td>
</tr>
<!-- Society & Culture -->
<tr>
<td rowspan="5"><b>Society & Culture</b></td>
<td><b>K-Refer*</b></td>
<td align="center">64.0</td>
<td align="center">53.6</td>
<td align="center">66.4</td>
<td align="center">71.6</td>
<td align="center">72.4</td>
<td align="center">43.2</td>
<td align="center"><b>89.6</b></td>
</tr>
<tr>
<td><b>K-Refer-Hard*</b></td>
<td align="center">67.1</td>
<td align="center">42.9</td>
<td align="center">61.4</td>
<td align="center">69.3</td>
<td align="center">65.7</td>
<td align="center">36.4</td>
<td align="center"><b>86.4</b></td>
</tr>
<tr>
<td><b>Ko-Sovereign*</b></td>
<td align="center">44.4</td>
<td align="center">35.8</td>
<td align="center">36.7</td>
<td align="center">46.9</td>
<td align="center"><b>49.8</b></td>
<td align="center">33.8</td>
<td align="center">56.3</td>
</tr>
<tr>
<td><b>HAERAE*</b></td>
<td align="center">61.3</td>
<td align="center">50.6</td>
<td align="center">70.8</td>
<td align="center">72.9</td>
<td align="center">68.4</td>
<td align="center">49.5</td>
<td align="center"><b>81.5</b></td>
</tr>
<tr>
<td><b>Avg.</b></td>
<td align="center">59.2</td>
<td align="center">45.7</td>
<td align="center">58.8</td>
<td align="center">65.2</td>
<td align="center">64.1</td>
<td align="center">40.7</td>
<td align="center"><b>78.4</b></td>
</tr>
<!-- Reasoning (Domain) -->
<tr>
<td rowspan="3"><b>Reasoning (Domain)</b></td>
<td><b>KMMLU</b></td>
<td align="center">43.5</td>
<td align="center">50.6</td>
<td align="center">45.1</td>
<td align="center">52.6</td>
<td align="center">55.4</td>
<td align="center">33.0</td>
<td align="center"><b>57.3</b></td>
</tr>
<tr>
<td><b>Ko-Sovereign*</b></td>
<td align="center">42.4</td>
<td align="center">42.5</td>
<td align="center">42.4</td>
<td align="center">45.6</td>
<td align="center">54.7</td>
<td align="center">36.7</td>
<td align="center"><b>58.0</b></td>
</tr>
<tr>
<td><b>Avg.</b></td>
<td align="center">43.0</td>
<td align="center">46.5</td>
<td align="center">43.8</td>
<td align="center">49.1</td>
<td align="center">55.1</td>
<td align="center">34.8</td>
<td align="center"><b>57.7</b></td>
</tr>
<!-- Instruction Following -->
<tr>
<td rowspan="3"><b>Instruction Following</b></td>
<td><b>Ko-IFEval*</b></td>
<td align="center">65.4</td>
<td align="center">75.9</td>
<td align="center">73.3</td>
<td align="center">69.1</td>
<td align="center"><b>83.6</b></td>
<td align="center">60.1</td>
<td align="center">82.0</td>
</tr>
<tr>
<td><b>Ko-MTBench</b></td>
<td align="center">74.0</td>
<td align="center">63.0</td>
<td align="center">74.0</td>
<td align="center">79.6</td>
<td align="center">71.0</td>
<td align="center">57.0</td>
<td align="center"><b>89.7</b></td>
</tr>
<tr>
<td><b>Avg.</b></td>
<td align="center">68.9</td>
<td align="center">69.4</td>
<td align="center">73.6</td>
<td align="center">74.4</td>
<td align="center">77.3</td>
<td align="center">58.5</td>
<td align="center"><b>85.9</b></td>
</tr>
</tbody>
</table>
`*` indicates KT proprietary evaluation resources.
<br>
# Usage
### Run on Friendli.AI
You can try our model immediately via `Friendli.AI`. Simply click `Deploy` and then `Friendli Endpoints`.
> [!Note]
> Please note that a login to `Friendli.AI` is required after your fifth chat interaction.
<p>
<img src="./assets/image_1.png" alt="Left Image" width="36%" style="display:inline-block; margin-right:2%">
<img src="./assets/image_2.png" alt="Right Image" width="36%" style="display:inline-block">
</p>
### Run on Your Local Machine
We provide a detailed description about running Mi:dm 2.0 on your local machine using llama.cpp, LM Studio, and Ollama. Please check our [github]() for more information
### Deployment
To serve Mi:dm 2.0 using [vLLM](https://github.com/vllm-project/vllm)(`>=0.8.0`) with an OpenAI-compatible API:
```bash
vllm serve K-intelligence/Midm-2.0-Mini-Instruct
```
### Tutorials
To help our end-users easily use Mi:dm 2.0, we have provided comprehensive tutorials on [github]().
<br>
<br>
<br>
# More Information
### Limitation
* The training data for both Mi:dm 2.0 models consists primarily of English and Korean. Understanding and generation in other languages are not guaranteed.
* The model is not guaranteed to provide reliable advice in fields that require professional expertise, such as law, medicine, or finance.
* Researchers have made efforts to exclude unethical content from the training data — such as profanity, slurs, bias, and discriminatory language. However, despite these efforts, the model may still produce inappropriate expressions or factual inaccuracies.
### License
Mi:dm 2.0 is licensed under the [MIT License](./LICENSE).
<!-- ### Citation
```
@misc{,
title={},
author={},
year={2025},
eprint={},
archivePrefix={arXiv},
primaryClass={cs.CL},
url={},
}
``` -->
### Contact
- Mi:dm 2.0 Technical Inquiries: [email protected]
<br>