# Export ONNX model to a LiteRT

You can convert ONNX models to LiteRT models and optimize them for on-device inference.

Converting an ONNX model to TensorFlow Lite (TFLite) is a common workflow, but it’s not a one-step conversion. The usual path is:

ONNX > TensorFlow (SavedModel) > TFLite

## Convert a ONNX model to a TensorFlow model

Use the `onnx-tf` module to convert an ONNX model to a TensorFlow Model. It’s the commonly used and stable approach.

1. Install dependencies.

pip install onnx onnx-tf tensorflow
        Copy to clipboard
2. Convert the model.

onnx_model_path=my_model.onnx
        Copy to clipboard

tf_model_path=tf_model
        Copy to clipboard

onnx-tf convert -i ${onnx_model_path} \
                        -o ${tf_model_path}
        Copy to clipboard

## Convert TensorFlow to LiteRT

import tensorflow as tf
    
    converter = tf.lite.TFLiteConverter.from_saved_model("tf_model")
    
    tflite_model = converter.convert()
    
    with open("model.tflite", "wb") as f:
        f.write(tflite_model)
    Copy to clipboard

## Quantization

See [quantize models using full integer quantization](https://docs.qualcomm.com/doc/80-80022-15B/topic/export-tf-model-litert.html#full-integer-quantization)
to quantize the model.

Last Published: Jun 23, 2026

[Previous Topic
Export a TensorFlow model to LiteRT](https://docs.qualcomm.com/bundle/publicresource/80-80022-15B/topics/export-tf-model-litert.md) [Next Topic
Export a Pytorch model to LiteRT](https://docs.qualcomm.com/bundle/publicresource/80-80022-15B/topics/export-pytorch-model-litert.md)