Instructions to use Tiranyx/migancore-0.14 with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Notebooks
Google Colab
Kaggle
Local Apps Settings

How to use Tiranyx/migancore-0.14 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 Tiranyx/migancore-0.14:Q4_K_M
# Run inference directly in the terminal:
llama cli -hf Tiranyx/migancore-0.14:Q4_K_M

Install from WinGet (Windows)

winget install llama.cpp
# Start a local OpenAI-compatible server with a web UI:
llama serve -hf Tiranyx/migancore-0.14:Q4_K_M
# Run inference directly in the terminal:
llama cli -hf Tiranyx/migancore-0.14:Q4_K_M

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 Tiranyx/migancore-0.14:Q4_K_M
# Run inference directly in the terminal:
./llama-cli -hf Tiranyx/migancore-0.14:Q4_K_M

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 Tiranyx/migancore-0.14:Q4_K_M
# Run inference directly in the terminal:
./build/bin/llama-cli -hf Tiranyx/migancore-0.14:Q4_K_M

Use Docker

docker model run hf.co/Tiranyx/migancore-0.14:Q4_K_M

LM Studio
Jan

vLLM

How to use Tiranyx/migancore-0.14 with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "Tiranyx/migancore-0.14"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "Tiranyx/migancore-0.14",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/Tiranyx/migancore-0.14:Q4_K_M

Ollama
How to use Tiranyx/migancore-0.14 with Ollama:
```
ollama run hf.co/Tiranyx/migancore-0.14:Q4_K_M
```
Unsloth Desktop

How to use Tiranyx/migancore-0.14 with Pi:

Start the llama.cpp server

# Install llama.cpp:
brew install llama.cpp
# Start a local OpenAI-compatible server:
llama serve -hf Tiranyx/migancore-0.14:Q4_K_M

Configure the model in Pi

# Install Pi:
npm install -g @earendil-works/pi-coding-agent
# Add to ~/.pi/agent/models.json:
{
  "providers": {
    "llama-cpp": {
      "baseUrl": "http://localhost:8080/v1",
      "api": "openai-completions",
      "apiKey": "none",
      "models": [
        {
          "id": "Tiranyx/migancore-0.14:Q4_K_M"
        }
      ]
    }
  }
}

Run Pi

# Start Pi in your project directory:
pi

Docker Model Runner
How to use Tiranyx/migancore-0.14 with Docker Model Runner:
```
docker model run hf.co/Tiranyx/migancore-0.14:Q4_K_M
```

Lemonade

How to use Tiranyx/migancore-0.14 with Lemonade:

Pull the model

# Download Lemonade from https://lemonade-server.ai/
lemonade pull Tiranyx/migancore-0.14:Q4_K_M

Run and chat with the model

lemonade run user.migancore-0.14-Q4_K_M

List all available models

lemonade list

Hermes Agent

How to use Tiranyx/migancore-0.14 with Hermes Agent:

Start the llama.cpp server

# Install llama.cpp:
brew install llama.cpp
# Start a local OpenAI-compatible server:
llama serve -hf Tiranyx/migancore-0.14:Q4_K_M

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 Tiranyx/migancore-0.14:Q4_K_M

Run Hermes

hermes

Atomic Chat

OpenClaw

How to use Tiranyx/migancore-0.14 with OpenClaw:

Start the llama.cpp server

# Install llama.cpp:
brew install llama.cpp
# Start a local OpenAI-compatible server:
llama serve -hf Tiranyx/migancore-0.14:Q4_K_M

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 "Tiranyx/migancore-0.14:Q4_K_M" \
  --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"

MiganCore 0.14 — open-source Indonesian LLM (archived research model, GGUF for Ollama)

migancore:0.14 was the served model of MiganCore, a one-person research project (May–September 2026). The project tried to give a small Indonesian model one ability: knowing where its own knowledge ends, and stopping there. The project closed on 28 September 2026. These weights are published as part of its open research record.

Research use only. This model fabricates on roughly half of the questions it should decline (see Measured behaviour). Do not use it to answer factual questions.

What it is

Base: Qwen/Qwen3-4B-Instruct-2507 (Apache-2.0).
Method: two LoRA adapters (r = 16, α = 32, 2 epochs), combined with a TIES merge (density 0.5, weights 1.0 / 1.0), then quantized to GGUF Q4_K_M.
- an arithmetic discipline adapter (285 rows);
- a style discipline adapter (200 rows).
Served: 24 August – 28 September 2026, on CPU through Ollama.

Training data (485 rows)

All rows were produced by deterministic program generators. No model wrote, rewrote, or judged any training row. Numbers are computed by code and re-verified mechanically. The generator code, including its answer templates, was written with the help of an AI coding assistant.

Adapter	Rows	Composition
Arithmetic	285	201 basic arithmetic · 76 unit calculations · 6 canary rows · 2 rows corrected by hand
Style	200	160 refusal patterns · 30 agent-format rows · 9 seed refusals · 1 canary row

Dataset fingerprints (sha256, first 16 hex): arithmetic 8009c9553fbb1261, style b2cdfc0336aa03a4.

Measured behaviour

All numbers are pre-registered verdicts, measured on the Indonesian-first petak-jujur2 battery (36 questions) on CPU.

Condition	Fabrication	Over-refusal	Fact accuracy	Valid rounds
Plain	49.9 %	3.1 %	0.43	40
With the abstention gate	35.8 %	1.3 %	0.43	28

Fabrication is the share of must-abstain questions on which the model asserted made-up content.
Over-refusal is the share of answerable, factual questions it declined.
Against its own base. On 25 Sep 2026, with identical requests, this model fabricated on 53.7 % of must-abstain questions, against 14.7 % for its base. The pre-registered anti-evasion guard failed, so no honesty claim is made either way.
Identity. The model may answer as its base model when asked who it is.

Files

File	What
`migancore-0.14-q4_k_m.gguf`	Merged model, GGUF Q4_K_M (2,497,278,784 bytes)
`adapters/lora-hitung-promptragam.tgz`	Arithmetic adapter
`adapters/lora-gaya.tgz`	Style adapter
`kemas_merge14.py`	The TIES merge recipe that produced the merged weights
`Modelfile`	Ollama definition used when it was served (`num_ctx 4096`, `temperature 0.3`)

Use it with Ollama:

ollama create migancore-0.14 -f Modelfile

Code, instruments, experiments and lineage: github.com/fahmiwol/migancore
Method and closing report: github.com/fahmiwol/migancore-research-method
Research record dataset: Tiranyx/migancore-research-record

License: Apache-2.0, following the base model. Author: Fahmi Ghani (fahmiwol@gmail.com).

Downloads last month: 31

GGUF

Model size

4B params

Architecture

qwen3

Hardware compatibility

4-bit

Model tree for Tiranyx/migancore-0.14

Base model

Qwen/Qwen3-4B-Instruct-2507

Adapter

(5791)

this model

Collection including Tiranyx/migancore-0.14

MiganCore: open-source Indonesian LLM research (2026)

Collection

Model, research record and explorer of MiganCore, an open-source Indonesian LLM project on hallucination and abstention. Closed Sep 2026. • 3 items • Updated 10 days ago

Tiranyx
/

migancore-0.14

MiganCore 0.14 — open-source Indonesian LLM (archived research model, GGUF for Ollama)

What it is

Training data (485 rows)

Measured behaviour

Files

More

Model tree for Tiranyx/migancore-0.14

Collection including Tiranyx/migancore-0.14

MiganCore: open-source Indonesian LLM research (2026)