# [Add an LLM to your Android app with Arm's AI Chat library](https://learn.arm.com/learning-paths/mobile-graphics-and-gaming/android-ai-chat-lib/)

## In this learning path

- [Introduction](https://learn.arm.com/learning-paths/mobile-graphics-and-gaming/android-ai-chat-lib/)
- [Create the Android project](https://learn.arm.com/learning-paths/mobile-graphics-and-gaming/android-ai-chat-lib/setup-1/)
- [Configure the AI Chat library dependency](https://learn.arm.com/learning-paths/mobile-graphics-and-gaming/android-ai-chat-lib/add-ai-chat-lib-2/)
- [Create the UI layouts and message adapter](https://learn.arm.com/learning-paths/mobile-graphics-and-gaming/android-ai-chat-lib/add-layouts-adapter-3/)
- [Implement the main activity logic](https://learn.arm.com/learning-paths/mobile-graphics-and-gaming/android-ai-chat-lib/use-lib-in-code-4/)
- [Download a model and run the app](https://learn.arm.com/learning-paths/mobile-graphics-and-gaming/android-ai-chat-lib/add-model-run-5/)
- [Next Steps](https://learn.arm.com/learning-paths/mobile-graphics-and-gaming/android-ai-chat-lib/_next-steps/)

## About this Learning Path

|                       |                       |
|-----------------------|-----------------------|
| Skill level:          | Introductory          |
| Reading time:         | 15 min                |
| Last updated:         | 05 Aug 2026           |

|                       |                       |
|-----------------------|-----------------------|
| Author:               | Ben Clark, Arm        |
| Arm IP:               | [Cortex-A](https://support.arm.com/?tab=compute-ip&Product%20Type=Application%20Processors) |
| Tags:                 | ML, Android, Kotlin, Neon, SVE2, SME2, LLM |

### Who is this for?
This is an introductory topic for developers who want to add a local, on-device LLM chat experience using Arm's AI Chat library, Kotlin, and Android Studio.

### What will you learn?
Upon completion of this Learning Path, you will be able to:
- Create a simple Android chatbot app scaffold in Android Studio
- Load a mobile-friendly GGUF model on-device and run streamed chat inference

### Prerequisites
Before starting, you will need the following:
- An Android development environment with Android Studio installed
- An Android phone for testing, in Developer Mode, with USB cable for connection
- Basic familiarity with Kotlin and Android app development

### Summary
You’ll build a minimal Android chatbot that runs a local `GGUF` model with Arm’s AI Chat library. First, you’ll configure the Android Studio project, add the Maven dependency, and create a chat UI. Then, you’ll connect the `MainActivity` class to load a mobile-friendly model and stream responses. Finally, you’ll download a `GGUF` model file on the device and run the app to verify on-device output.

### Frequently asked questions

<details>
<summary>Which Gradle file should I edit to add the AI Chat dependency?</summary>
Add the dependency in the app module’s `build.gradle.kts`, not the project-level `build.gradle.kts`. Place the implementation for `com.arm:ai-chat:0.1.0` in the dependencies block.
</details>

<details>
<summary>How do I check my repository configuration so the library resolves?</summary>
Open `settings.gradle.kts` and confirm the top-level repositories include `google()` and `mavenCentral()`. With both present, you can resolve the AI Chat library from Maven Central.
</details>

<details>
<summary>Which layout file do I replace for the chat UI?</summary>
Replace `activity_main.xml` in `app/src/main/res/layout` with the provided XML. The XML defines a status area, a message list, and a text input with a send button.
</details>

<details>
<summary>How do I choose a GGUF model that fits my device?</summary>
Select a model that is significantly smaller than your phone’s available RAM. A mobile-friendly example is `google_gemma-3-4b-it-Q4_0.gguf`, a Q4_0-quantized Gemma 3 4B model that works well with Arm’s KleidiAI and benefits devices with SME2, SVE2, or Neon capabilities.
</details>

<details>
<summary>What result should I expect when I run the app?</summary>
After you provide the `GGUF` file on the device, you should see the app load the model and stream chat responses into the message list. Check the status area during generation to confirm progress.
</details>