Intelligence, uncompromised.

A serious environment for running language models directly on Android.

01 — LOCAL INFERENCE

Run compatible AI models directly on Android.

Moonlight brings the power of `llama.cpp` to your pocket. Connect with local GGUF models running natively on your device CPU and GPU.

  • Support for 4-bit and 8-bit quantization
  • Zero internet required for generation
9:41
GGUF

llama.cpp Engine

Native Android backend compiling the model directly into memory.

02 — MODEL STORE

Choose your intelligence.

Download verified models directly from Hugging Face. The Model Store manages your local catalog safely.

  • SHA-256 verification on download
  • Automatic storage management
9:41
Model Catalog
Search models...
Installed
P3
Phi-3 Mini
2.2 GB • Offline Ready
How model installation works
D1
DeepSeek R1 7B
Verifying SHA-256 Checksum...
Wait
Download
Verify
Install
Available Models
G2
Gemma 2 9B
5.4 GB • Google
Q2
Qwen 2 7B
4.4 GB • Alibaba
L1
LLaVA 1.6 7B
4.8 GB • LLaVA Team
03 — PRIVATE CONVERSATIONS

A beautiful interface for your ideas.

Interact with your models using an elegant, fluid chat interface designed for Android. Everything is saved strictly on-device.

9:41
M
Phi-3 Mini
Running locally ? 19.4 t/s
Can you explain how local LLM inference works without the internet?
M

Local LLM inference runs on your device's hardware (CPU/GPU) rather than sending data to a cloud server.

Moonlight uses llama.cpp which is highly optimized for mobile devices. The model weights are stored locally on your phone's storage in a quantized format (GGUF).

Because the computation happens on-device, it works offline and your chat history is stored locally.

Message Moonlight...

In Development

Features currently being built in the Moonlight repository. Not yet available in the production APK.

Roadmap

Biometric Security

Ensure your private AI conversations remain private even if you hand your unlocked phone to someone else.

Roadmap

Local Memory

Moonlight is currently wiring up its backend MemoryEngine. Soon, it will extract facts from your conversations and store them in a local vector database.

Roadmap

Semantic Search

The SemanticSearchEngine is being integrated into the UI. It will allow you to find past conversations based on meaning.

Roadmap

Multimodal Vision

Integration for Vision models (like LLaVA) is in development. You will be able to share images with Moonlight for local analysis.

Roadmap

OCR Document Scanning

The CameraEngine is being wired up to allow on-device text extraction from photos and documents.

Roadmap

Autonomous Agents

The AgentEngine is being implemented to allow models to execute multi-step reasoning and use tools.

Roadmap

Workflows

The WorkflowEngine will allow you to create and save multi-step prompt chains.

Roadmap

Local Plugins

The PluginSandbox is being built to allow secure, local extensions to Moonlight's functionality.

Roadmap

Knowledge Graph

A local Knowledge Graph is in development to link extracted entities and facts together.