🤗 Mi:dm 2.0 Models | 📜 Mi:dm 2.0 Technical Report | 📕 Mi:dm 2.0 Technical Blog*

_{*To be released soon}

• 🔜 (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🤗.

• Overview
  • Mi:dm 2.0
  • Quickstart
  • Evaluation
• Usage
  • Run on Your Local Machine
    • llama.cpp
    • LM Studio
    • Ollama
  • Deployment
  • Tutorials
• More Information
  • Limitation
  • License
  • Contact

Mi:dm 2.0 is a "Korea-centric AI" model developed using KT's proprietary technology. The term "Korea-centric AI" refers to a model that deeply internalizes the unique values, cognitive frameworks, and commonsense reasoning inherent to Korean society. It goes beyond simply processing or generating Korean text—it reflects a deeper understanding of the socio-cultural norms and values that define Korean society.

Mi:dm 2.0 is released in two versions:

• Mi:dm 2.0 Base
  An 11.5B parameter dense model designed to balance model size and performance.
  It extends an 8B-scale model by applying the Depth-up Scaling (DuS) method, making it suitable for real-world applications that require both performance and versatility.

• Mi:dm 2.0 Mini
  A lightweight 2.3B parameter dense model optimized for on-device environments and systems with limited GPU resources. It was derived from the Base model through pruning and distillation to enable compact deployment.

Here is the code snippet to run conversational inference with the model:

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]))

* indicates KT proprietary evaluation resources.

We assume that you have installed llama.cpp.

1. Download Mi:dm 2.0 model in GGUF format from our Hugging Face repository.

   huggingface-cli download K-intelligence/Midm-2.0-Base-Instruct-GGUF \
       --include "Midm-2.0-Base-Instruct-GGUF-BF16.gguf" \
       --local-dir .
   

2. To run the model, use the command below:

   llama-cli

   MODEL_PATH=./Midm-2.0-Base-Instruct-GGUF-BF16.gguf
   
   llama-cli \
       -m ${MODEL_PATH} \
       -p "KT에 대해 소개해줘" \ # customize your prompt here
       --temp 0.7
   

   llama-server

   python3 -m llama_cpp.server \
     --model ${MODEL_PATH} \
   

You can easily use Mi:dm 2.0 in LM Studio with just a few clicks.

1. Install LM Studio.
2. In Discover tab of LM Studio, search for and download the model. Alternatively, you can manually download Mi:dm 2.0 model in GGUF format using huggingface-cli.

   huggingface-cli download K-intelligence/Midm-2.0-Base-Instruct-GGUF \
       --include "Midm-2.0-Base-Instruct-GGUF-BF16.gguf" \
       --local-dir .

   If you downloaded the model manually, move the file to the LM Studio model directory (~/.lmstudio/models/) with the following structure:

   ~/.lmstudio/models/
   └── midm/
       └── midm-2.0-base/
           └── Midm-2.0-Base-Instruct-GGUF-BF16.gguf
   

3. In Chat tab, run the model in conversational mode.
4. To use OpenAI-compatible API endpoints, start the server from Developer tab.
5. To connect Mi:dm 2.0 with the MCP (Model Context Protocol) server for tool usage, go to Chat tab in LM Studio and update Program settings by adding or modifying the mcp.json file as follows:

      {
        "mcpServers": {
          "server-time": {
            "command": "uvx",
            "args": [
              "mcp-server-time",
              "--local-timezone=Asia/Seoul"
            ]
          },
          "ddg-search": {
            "command": "uvx",
            "args": [
              "duckduckgo-mcp-server"
            ]
          }
        }
      }

We provide a guide for running Mi:dm 2.0 via Ollama in our Open WebUI tutorial. Once the setup is complete, you can run Mi:dm 2.0 using the following command:

ollama run midm-2.0:base

To serve Mi:dm 2.0 using vLLM(>=0.8.0) with an OpenAI-compatible API:

vllm serve K-intelligence/Midm-2.0-Base-Instruct

To help our end-users easily use Mi:dm 2.0, we have provided comprehensive tutorials on our tutorial page. We provide several examples to help you get started with Mi:dm 2.0 quickly and easily:

• 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.

Mi:dm 2.0 is licensed under the MIT License.

Mi:dm 2.0 Technical Inquiries: [email protected]