Instructions to use midnightcoderagent/MidnightCoder-30B with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use midnightcoderagent/MidnightCoder-30B with llama-cpp-python:

# !pip install llama-cpp-python

from llama_cpp import Llama

llm = Llama.from_pretrained(
	repo_id="midnightcoderagent/MidnightCoder-30B",
	filename="MidnightCoder-30B.gguf",
)

llm.create_chat_completion(
	messages = [
		{
			"role": "user",
			"content": "What is the capital of France?"
		}
	]
)

Notebooks
Google Colab
Kaggle
Local Apps Settings

llama.cpp

How to use midnightcoderagent/MidnightCoder-30B with llama.cpp:

Install (macOS, Linux)

curl -LsSf https://llama.app/install.sh | sh
# Start a local OpenAI-compatible server with a web UI:
llama serve -hf midnightcoderagent/MidnightCoder-30B
# Run inference directly in the terminal:
llama cli -hf midnightcoderagent/MidnightCoder-30B

Install from WinGet (Windows)

winget install llama.cpp
# Start a local OpenAI-compatible server with a web UI:
llama serve -hf midnightcoderagent/MidnightCoder-30B
# Run inference directly in the terminal:
llama cli -hf midnightcoderagent/MidnightCoder-30B

Use pre-built binary

# Download pre-built binary from:
# https://github.com/ggerganov/llama.cpp/releases
# Start a local OpenAI-compatible server with a web UI:
./llama-server -hf midnightcoderagent/MidnightCoder-30B
# Run inference directly in the terminal:
./llama-cli -hf midnightcoderagent/MidnightCoder-30B

Build from source code

git clone https://github.com/ggerganov/llama.cpp.git
cd llama.cpp
cmake -B build
cmake --build build -j --target llama-server llama-cli
# Start a local OpenAI-compatible server with a web UI:
./build/bin/llama-server -hf midnightcoderagent/MidnightCoder-30B
# Run inference directly in the terminal:
./build/bin/llama-cli -hf midnightcoderagent/MidnightCoder-30B

Use Docker

docker model run hf.co/midnightcoderagent/MidnightCoder-30B

LM Studio
Jan

vLLM

How to use midnightcoderagent/MidnightCoder-30B with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "midnightcoderagent/MidnightCoder-30B"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "midnightcoderagent/MidnightCoder-30B",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/midnightcoderagent/MidnightCoder-30B

Ollama
How to use midnightcoderagent/MidnightCoder-30B with Ollama:
```
ollama run hf.co/midnightcoderagent/MidnightCoder-30B
```

Unsloth Studio

How to use midnightcoderagent/MidnightCoder-30B with Unsloth Studio:

Install Unsloth Studio (macOS, Linux, WSL)

curl -fsSL https://unsloth.ai/install.sh | sh
# Run unsloth studio
unsloth studio -H 0.0.0.0 -p 8888
# Then open http://localhost:8888 in your browser
# Search for midnightcoderagent/MidnightCoder-30B to start chatting

Install Unsloth Studio (Windows)

irm https://unsloth.ai/install.ps1 | iex
# Run unsloth studio
unsloth studio -H 0.0.0.0 -p 8888
# Then open http://localhost:8888 in your browser
# Search for midnightcoderagent/MidnightCoder-30B to start chatting

Using HuggingFace Spaces for Unsloth

# No setup required
# Open https://huggingface.co/spaces/unsloth/studio in your browser
# Search for midnightcoderagent/MidnightCoder-30B to start chatting

How to use midnightcoderagent/MidnightCoder-30B with Pi:

Start the llama.cpp server

# Install llama.cpp:
brew install llama.cpp
# Start a local OpenAI-compatible server:
llama serve -hf midnightcoderagent/MidnightCoder-30B

Configure the model in Pi

# Install Pi:
npm install -g @mariozechner/pi-coding-agent
# Add to ~/.pi/agent/models.json:
{
  "providers": {
    "llama-cpp": {
      "baseUrl": "http://localhost:8080/v1",
      "api": "openai-completions",
      "apiKey": "none",
      "models": [
        {
          "id": "midnightcoderagent/MidnightCoder-30B"
        }
      ]
    }
  }
}

Run Pi

# Start Pi in your project directory:
pi

Hermes Agent new

How to use midnightcoderagent/MidnightCoder-30B with Hermes Agent:

Start the llama.cpp server

# Install llama.cpp:
brew install llama.cpp
# Start a local OpenAI-compatible server:
llama serve -hf midnightcoderagent/MidnightCoder-30B

Configure Hermes

# Install Hermes:
curl -fsSL https://hermes-agent.nousresearch.com/install.sh | bash
hermes setup
# Point Hermes at the local server:
hermes config set model.provider custom
hermes config set model.base_url http://127.0.0.1:8080/v1
hermes config set model.default midnightcoderagent/MidnightCoder-30B

Run Hermes

hermes

Atomic Chat new

OpenClaw new

How to use midnightcoderagent/MidnightCoder-30B with OpenClaw:

Start the llama.cpp server

# Install llama.cpp:
brew install llama.cpp
# Start a local OpenAI-compatible server:
llama serve -hf midnightcoderagent/MidnightCoder-30B

Configure OpenClaw

# Install OpenClaw:
npm install -g openclaw@latest
# Register the local server and set it as the default model:
openclaw onboard --non-interactive --mode local \
  --auth-choice custom-api-key \
  --custom-base-url http://127.0.0.1:8080/v1 \
  --custom-model-id "midnightcoderagent/MidnightCoder-30B" \
  --custom-provider-id llama-cpp \
  --custom-compatibility openai \
  --custom-text-input \
  --accept-risk \
  --skip-health

Run OpenClaw

openclaw agent --local --agent main --message "Hello from Hugging Face"

Docker Model Runner
How to use midnightcoderagent/MidnightCoder-30B with Docker Model Runner:
```
docker model run hf.co/midnightcoderagent/MidnightCoder-30B
```

Lemonade

How to use midnightcoderagent/MidnightCoder-30B with Lemonade:

Pull the model

# Download Lemonade from https://lemonade-server.ai/
lemonade pull midnightcoderagent/MidnightCoder-30B

Run and chat with the model

lemonade run user.MidnightCoder-30B-{{QUANT_TAG}}

List all available models

lemonade list

🌙 MidnightCoder-30B-GGUF

A GGUF distribution of Qwen3-Coder-30B-A3B-Instruct optimized for the Midnight Coder software engineering agent.

MidnightCoder-30B-GGUF is designed for developers who want to run a powerful coding model locally using llama.cpp, Ollama, LM Studio, or any GGUF-compatible inference engine.

This distribution is optimized for the Midnight Coder agent but is fully compatible with any coding agent or workflow. It excels at structured, specification-driven software engineering while remaining suitable for general-purpose coding tasks.

🔗 Midnight Coder Agent

GitHub: https://github.com/midnightcoderagent/Midnight-Coder
Website: https://midnightcoderagent.github.io
Install: npm install -g midnight-coder (Linux support currently available. Windows and macOS support coming soon.)
Issues & Feature Requests: https://github.com/midnightcoderagent/Midnight-Coder/issues

Features

🚀 Optimized for software engineering
📝 Specification-driven development
🤖 Designed for AI coding agents
🔧 Compatible with llama.cpp
🦙 Compatible with Ollama
💻 Local-first development
📚 Strong support for large codebases
🔍 Refactoring and debugging assistance
🧪 Test generation support
📖 Documentation generation

Base Model

This model is based on:

Qwen3-Coder-30B-A3B-Instruct

All credit for the pretrained model belongs to the Qwen Team (Alibaba).

This repository distributes the model in GGUF format for local inference.

About Midnight Coder

Midnight Coder is an open-source AI coding assistant focused on structured software engineering.

Instead of behaving like a generic chatbot, Midnight Coder follows an engineering workflow centered around planning, specifications, implementation, and verification.

Typical workflow:

Understand the task
Create a technical specification
Define implementation scope
Implement changes
Verify results
Report completed work

This workflow helps produce predictable, maintainable, and review-friendly code.

Intended Use

MidnightCoder-30B-GGUF is suitable for:

Software Engineering
Backend Development
Frontend Development
API Design
SQL
Docker
DevOps
Linux
Code Review
Refactoring
Documentation
Debugging
Architecture Discussions

Compatible Software

The model can be used with:

llama.cpp
Ollama
LM Studio
Open WebUI
Jan
KoboldCpp
Text Generation WebUI
Any GGUF-compatible runtime

Recommended Context

Recommended context sizes:

Context	Recommended Usage
8K	Small projects
16K	General development
32K	Large repositories
64K+	Complex multi-file projects (hardware permitting)

Quantization

This repository provides the model in GGUF format.

The quantization is indicated in the filename.

Examples:

Q4_K_M
Q5_K_M
Q6_K
Q8_0

Example (llama.cpp)

llama-cli \
-m MidnightCoder-30B-Q4_K_M.gguf \
-c 32768

Example (Ollama)

Create a Modelfile:

FROM MidnightCoder-30B-Q4_K_M.gguf

SYSTEM """
<Midnight Coder system prompt>
"""

Build:

ollama create midnightcoder -f Modelfile

Run:

ollama run midnightcoder

Performance Notes

The required memory depends on:

Quantization
Context size
KV cache precision
Runtime configuration

Lower quantizations reduce memory usage while higher quantizations generally preserve more model quality.

Limitations

Like any language model:

Responses may contain inaccuracies.
Generated code should always be reviewed.
Security-sensitive code requires human validation.
Production deployments should include proper testing.

License

This repository distributes a GGUF version of the original model.

Please refer to the original Qwen3-Coder-30B-A3B-Instruct license for licensing terms regarding the base model.

Acknowledgements

Special thanks to:

Alibaba Qwen Team
Hugging Face
llama.cpp
Ollama

for making local AI development accessible to the community.

Midnight Coder

Midnight Coder is an open-source project dedicated to building high-quality AI tools for software developers.

Links

GitHub

https://github.com/midnightcoderagent/Midnight-Coder

Organization

https://huggingface.co/MidnightCoder

Disclaimer

This repository is an independent GGUF distribution based on Qwen3-Coder-30B-A3B-Instruct.

It is not an official release from the Qwen Team or Alibaba.

All trademarks and original model rights belong to their respective owners.

Downloads last month: 251

GGUF

Model size

31B params

Architecture

qwen3moe

Hardware compatibility

We're not able to determine the quantization variants.

View all variants

Model tree for midnightcoderagent/MidnightCoder-30B

Base model

Qwen/Qwen3-Coder-30B-A3B-Instruct

Quantized

(151)

this model