Instructions to use immortaltatsu/ghost-ai-gguf with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use immortaltatsu/ghost-ai-gguf with llama-cpp-python:

# !pip install llama-cpp-python

from llama_cpp import Llama

llm = Llama.from_pretrained(
	repo_id="immortaltatsu/ghost-ai-gguf",
	filename="ghost-ai-Q6_K.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 immortaltatsu/ghost-ai-gguf with llama.cpp:

Install from brew

brew install llama.cpp
# Start a local OpenAI-compatible server with a web UI:
llama-server -hf immortaltatsu/ghost-ai-gguf:Q6_K
# Run inference directly in the terminal:
llama-cli -hf immortaltatsu/ghost-ai-gguf:Q6_K

Install from WinGet (Windows)

winget install llama.cpp
# Start a local OpenAI-compatible server with a web UI:
llama-server -hf immortaltatsu/ghost-ai-gguf:Q6_K
# Run inference directly in the terminal:
llama-cli -hf immortaltatsu/ghost-ai-gguf:Q6_K

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 immortaltatsu/ghost-ai-gguf:Q6_K
# Run inference directly in the terminal:
./llama-cli -hf immortaltatsu/ghost-ai-gguf:Q6_K

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 immortaltatsu/ghost-ai-gguf:Q6_K
# Run inference directly in the terminal:
./build/bin/llama-cli -hf immortaltatsu/ghost-ai-gguf:Q6_K

Use Docker

docker model run hf.co/immortaltatsu/ghost-ai-gguf:Q6_K

LM Studio
Jan

vLLM

How to use immortaltatsu/ghost-ai-gguf with vLLM:

Install from pip and serve model

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

Use Docker

docker model run hf.co/immortaltatsu/ghost-ai-gguf:Q6_K

Ollama
How to use immortaltatsu/ghost-ai-gguf with Ollama:
```
ollama run hf.co/immortaltatsu/ghost-ai-gguf:Q6_K
```

Unsloth Studio

How to use immortaltatsu/ghost-ai-gguf 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 immortaltatsu/ghost-ai-gguf 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 immortaltatsu/ghost-ai-gguf to start chatting

Using HuggingFace Spaces for Unsloth

# No setup required
# Open https://huggingface.co/spaces/unsloth/studio in your browser
# Search for immortaltatsu/ghost-ai-gguf to start chatting

How to use immortaltatsu/ghost-ai-gguf with Pi:

Start the llama.cpp server

# Install llama.cpp:
brew install llama.cpp
# Start a local OpenAI-compatible server:
llama-server -hf immortaltatsu/ghost-ai-gguf:Q6_K

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": "immortaltatsu/ghost-ai-gguf:Q6_K"
        }
      ]
    }
  }
}

Run Pi

# Start Pi in your project directory:
pi

Hermes Agent new

How to use immortaltatsu/ghost-ai-gguf with Hermes Agent:

Start the llama.cpp server

# Install llama.cpp:
brew install llama.cpp
# Start a local OpenAI-compatible server:
llama-server -hf immortaltatsu/ghost-ai-gguf:Q6_K

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 immortaltatsu/ghost-ai-gguf:Q6_K

Run Hermes

hermes

Docker Model Runner
How to use immortaltatsu/ghost-ai-gguf with Docker Model Runner:
```
docker model run hf.co/immortaltatsu/ghost-ai-gguf:Q6_K
```

Lemonade

How to use immortaltatsu/ghost-ai-gguf with Lemonade:

Pull the model

# Download Lemonade from https://lemonade-server.ai/
lemonade pull immortaltatsu/ghost-ai-gguf:Q6_K

Run and chat with the model

lemonade run user.ghost-ai-gguf-Q6_K

List all available models

lemonade list

Ghost AI

The ultra-lean, on-device AI model behind Ghost AI on Solana.

Ghost AI is part of Ghost AI, an on-device privacy-first crypto application running on Solana. This model powers the smallest, fastest local AI surface inside the Ghost AI app — perfect for mobile and low-resource devices. Your prompts, wallet context, and financial data never leave your device.

Ghost AI is a fork of LiquidAI's LFM 2.5 (350M), rebranded and packaged for on-device crypto and finance use cases. All underlying model weights, architecture, and capabilities derive from LFM 2.5 — credit and licensing terms remain with the upstream project.

Ghost AI (350M) is the base / lite tier of the Ghost AI model family. For higher-quality reasoning, see Ghost AI Pro (1.2B, based on LFM 2.5 1.2B Instruct).

Why Ghost AI

Ultra-light. Q6_K quantization — ~280MB on disk. Runs comfortably on phones, embedded devices, and low-end laptops.
On-device, always private. Runs locally via llama.cpp, LM Studio, Ollama, or any GGUF-compatible runtime. Nothing is logged or sent to a server.
Built for finance + crypto. Wallet-aware prompts, on-chain transaction explanations, DeFi context, and personal-finance reasoning.
Native to Ghost AI on Solana. The default local model for the Ghost AI privacy-first crypto app on Solana.
Real-time. Designed for instant, low-latency responses on CPU — no GPU required.

Ghost AI model family

Tier	Repo	Base	Quantization	Size	Best for
Ghost AI (this)	`immortaltatsu/ghost-ai-gguf`	LFM 2.5 350M	Q6_K	~280 MB	Mobile, embedded, instant responses
Ghost AI Pro	`immortaltatsu/ghost-ai-pro-gguf`	LFM 2.5 1.2B Instruct	Q4_K_S	~700 MB	Desktop, deeper reasoning

Files

File	Quantization	Size	Notes
`ghost-ai-Q6_K.gguf`	Q6_K	~280 MB	Recommended. High-quality 6-bit quantization with minimal quality loss for a 350M model.

Usage

llama.cpp

./llama-cli -m ghost-ai-Q6_K.gguf -p "Explain this Solana transaction: ..." -n 256

Ollama (Modelfile)

FROM ./ghost-ai-Q6_K.gguf
PARAMETER temperature 0.7
PARAMETER num_ctx 4096

ollama create ghost-ai -f Modelfile
ollama run ghost-ai

LM Studio

Download the .gguf file and load it directly from the LM Studio model browser.

Intended Use

Ghost AI is intended for:

Powering the on-device AI layer of the Ghost AI crypto app on Solana
Private wallet, transaction, and portfolio explanations
Lightweight personal finance and budgeting assistance
DeFi term clarification and on-chain activity narration
Embedded assistants in mobile crypto and fintech apps where privacy is non-negotiable

Limitations

Ghost AI is a small (350M) model — fast and private, but limited in long-form reasoning compared to Ghost AI Pro. It is not a substitute for licensed financial advice, accounting, or legal counsel, and it is not a trading or transaction-signing agent. Verify any numerical reasoning, dates, ticker symbols, token addresses, on-chain claims, and regulatory statements against authoritative sources before acting on them. Always confirm wallet operations through the Ghost AI app's signing flow — never from raw model output.

About Ghost AI

Ghost AI is an on-device, privacy-first crypto app built on Solana. It brings local AI directly into the wallet experience — letting users explore transactions, balances, and DeFi context with a model that runs entirely on their own device. Nothing is logged, sent to a server, or shared with a third party.

This model is the lightweight default that powers Ghost AI's on-device intelligence layer on resource-constrained devices.

Upstream

Ghost AI is a fork of LiquidAI/LFM2.5-350M. The GGUF quantizations are sourced from LiquidAI/LFM2.5-350M-GGUF.

License

Released for research and on-device application use. This model is governed by the upstream LFM Open License issued by Liquid AI — please review the original terms before redistribution or commercial deployment.

Ghost AI — intelligence that stays with you.

Downloads last month: 156

GGUF

Model size

0.4B params

Architecture

lfm2

Hardware compatibility

6-bit

Model tree for immortaltatsu/ghost-ai-gguf

Base model

LiquidAI/LFM2.5-350M-Base

Finetuned

LiquidAI/LFM2.5-350M

Quantized

(33)

this model