Introducing Google AI Edge Portal: Benchmark Edge AI at scale. Sign-up to request access during private preview.
LiteRT for Android
The following LiteRT runtime APIs are available for Android development:
CompiledModelAPI: The modern standard for high-performance inference, streamlining hardware acceleration across CPU/GPU/NPU. Learn more about why to choose the CompiledModel API.InterpreterAPI: the basic inference API, maintained for backward compatibility.
Get Started with CompiledModel API
For classical ML models, see the following demo apps.
- Image segmentation Kotlin App: CPU/GPU/NPU inference.
- Image segmentation C++ App: CPU/GPU/NPU inference with async execution.
For GenAI models, see the following demo apps:
- EmbeddingGemma semantic similarity C++ App: CPU/GPU/NPU inference.
Supported Android Versions and APIs
| LiteRT Version | Status | Supported API | Min SDK Level | Min NDK version (if used) | Release Date |
|---|---|---|---|---|---|
v2.1.0 |
✅ Latest | CompiledModel Interpreter(CPU only) |
23 (Android 6 Marshmallow) |
r26a |
2025年12月19日 |
v2.0.3 |
⚠️ Legacy | CompiledModel |
26 (Android 8 Oreo) |
r26a |
2025年11月08日 |
v1.4.1 |
✅ Latest | Interpreter |
21 (Android 5 Lollipop) |
r26a |
2025年11月07日 |
v1.4.0 |
⚠️ Legacy | Interpreter |
26 (Android 8 Oreo) |
r26a |
2025年06月25日 |
v1.3.0 |
⚠️ Legacy | Interpreter |
21 (Android 5 Lollipop) |
r26a |
2025年05月19日 |
v1.2.0 |
⚠️ Legacy | Interpreter |
21 (Android 5 Lollipop) |
r26a |
2025年03月13日 |
Important: Keep your dependencies up to date to ensure compatibility with the latest features and security updates.
Quickstart with CompiledModel API
Add the LiteRT Maven package to your Android project:
dependencies {
...
implementation `com.google.ai.edge.litert:litert:2.1.0`
}
Integrate your .tflite model with the CompiledModel API. The following code
snippet shows the basic implementation in Kotlin and C++.
Kotlin
// Load model and initialize runtime
valcompiledModel=CompiledModel.create(
"/path/to/mymodel.tflite",
CompiledModel.Options(Accelerator.CPU))
// Preallocate input/output buffers
valinputBuffers=compiledModel.createInputBuffers()
valoutputBuffers=compiledModel.createOutputBuffers()
// Fill the input buffer
inputBuffers.get(0).writeFloat(input0)
inputBuffers.get(1).writeFloat(input1)
// Invoke
compiledModel.run(inputBuffers,outputBuffers)
// Read the output
valoutput=outputBuffers.get(0).readFloat()
C++
// Load model and initialize runtime
LITERT_ASSIGN_OR_RETURN(autoenv,GetEnvironment());
LITERT_ASSIGN_OR_RETURN(autooptions,GetOptions());
LITERT_ASSIGN_OR_RETURN(
autocompiled_model,
CompiledModel::Create(env,"/path/to/mymodel.tflite",options));
// Preallocate input/output buffers
LITERT_ASSIGN_OR_RETURN(autoinput_buffers,compiled_model.CreateInputBuffers(signature_index));
LITERT_ASSIGN_OR_RETURN(autooutput_buffers,compiled_model.CreateOutputBuffers(signature_index));
// Fill the input buffer
LITERT_ABORT_IF_ERROR(input_buffers[0].Write(input0));
LITERT_ABORT_IF_ERROR(input_buffers[1].Write(input1));
// Invoke
LITERT_ABORT_IF_ERROR(compiled_model.Run(signature_index,input_buffers,output_buffers));
// Read the output
LITERT_ABORT_IF_ERROR(output_buffers[0].Read(output0));