Open Source · Apache 2.0 · 100% On-Device

Private, On-Device AI Chat for Android

SmolChat lets you run Small Language Models (SLMs) in GGUF format directly on your Android device using llama.cpp—no cloud APIs, no internet required, and full control over prompts and inference parameters.

SmolChat screenshot 1 SmolChat screenshot 2 SmolChat screenshot 3 SmolChat screenshot 4 SmolChat screenshot 5 SmolChat screenshot 6

Built for Local SLM Inference

Designed around a clean Android interface and native C/C++ execution for fast, private conversations on your phone.

100% On-Device Execution

Runs GGUF models natively via llama.cpp compiled with the Android NDK. Your prompts and responses stay on your device.

Flexible GGUF Management

Add, remove, and switch between local GGUF models, or download models directly inside the app.

Custom Prompts & Sampling

Configure system prompts, temperature, and min-p per chat or set global defaults for all new conversations.

Downstream Tasks

Create reusable single-turn tasks for quick text generation without persisting chat history.

Markdown & Syntax Highlighting

Renders rich Markdown responses with Markwon and Prism4j, complete with one-tap copy for code blocks.

OpenAI-Compatible Server

Includes a local HTTP server capable of listing loaded models and serving OpenAI-compatible LLM inference.

App Screenshots

Explore the chat experience, model management, and inference settings on Android.

SmolChat screenshot 1
SmolChat screenshot 2
SmolChat screenshot 3
SmolChat screenshot 4
SmolChat screenshot 5
SmolChat screenshot 6

Installation

Choose the distribution channel that works best for you.

Google Play

Install directly from the Google Play Store with automatic updates.

GitHub Releases

  1. Download the latest APK from GitHub Releases.
  2. Transfer and install the APK on your Android device.

Obtainium

  1. Open Obtainium and tap Add App.
  2. Enter https://github.com/shubham0204/SmolChat-Android
  3. Receive release updates directly from GitHub.

Architecture

Visualizing how prompts flow from the Android UI down to native GGUF inference.

APP MODULE (KOTLIN / ANDROID UI) Chat & Tasks UI Markwon + Prism4j LLMModel & Chat DB Model path & persisted messages SMOLLM MODULE (KOTLIN + JNI + C++) SmolLM.kt Kotlin API smollm.cpp JNI Bindings llm_inference.cpp loadModel / addChatMessage NATIVE RUNTIME (ANDROID NDK) llama.cpp (C/C++ API) Local GGUF Model