---
title: LeViT
canonical_url: "https://www.modelscope.cn/models/Qualcomm-AI-Hub/LeViT"
md_url: "https://www.modelscope.cn/models/Qualcomm-AI-Hub/LeViT.md"
repository: Qualcomm-AI-Hub/LeViT
last_updated: 2026-10-16
license: apache-2.0
pipeline_tag: image-classification
tasks:
  - image-classification
downloads: 0
stars: 0
tags:
  - bu_auto
  - android
---

# LeViT

> LeViT - Qualcomm-AI-Hub 在 ModelScope 开源的模型。LeViT: Optimized for Qualcomm Devices

Qualcomm-AI-Hub/LeViT 是 ModelScope 魔搭社区上的image-classification模型，采用 apache-2.0 许可。

- **Repository**: Qualcomm-AI-Hub/LeViT
- **License**: apache-2.0
- **Tasks**: image-classification
- **Tags**: bu_auto, android
- **Downloads**: 0
- **Stars**: 0
- **Last updated**: 2026-10-16

Source: https://www.modelscope.cn/models/Qualcomm-AI-Hub/LeViT

---

![](https://china-qaihub-public-assets.s3.cn-northwest-1.amazonaws.com.cn/qai-hub-models/models/levit/web-assets/model_demo.png)

# LeViT: Optimized for Qualcomm Devices

LeViT is a vision transformer model that can classify images from the Imagenet dataset.

This is based on the implementation of LeViT found [here](https://github.com/facebookresearch/LeViT).
This repository contains pre-exported model files optimized for Qualcomm® devices. You can use the [Qualcomm® AI Hub Models](https://github.com/qualcomm/ai-hub-models-cn/blob/main/src/qai_hub_models/models/levit) library to export with custom configurations. More details on model performance across various devices, can be found [here](#performance-summary).

Qualcomm AI Hub Models uses [Qualcomm AI Hub Workbench](https://workbench.aihub.qualcomm.cn) to compile, profile, and evaluate this model. [Sign up](https://myaccount.qualcomm.com/signup) to run these models on a hosted Qualcomm® device.

## Getting Started
There are two ways to deploy this model on your device:

### Option 1: Download Pre-Exported Models

Below are pre-exported model assets ready for deployment.

| Runtime | Precision | Chipset | SDK Versions | Download |
|---|---|---|---|---|
| ONNX | float | Universal | QAIRT 2.45, ONNX Runtime 1.27.1 | [Download](https://china-qaihub-public-assets.s3.cn-northwest-1.amazonaws.com.cn/qai-hub-models/models/levit/releases/v0.62.1/levit-onnx-float.zip)
| ONNX | w8a16 | Universal | QAIRT 2.45, ONNX Runtime 1.27.1 | [Download](https://china-qaihub-public-assets.s3.cn-northwest-1.amazonaws.com.cn/qai-hub-models/models/levit/releases/v0.62.1/levit-onnx-w8a16.zip)
| ONNX | w8a16_mixed_int16 | Universal | QAIRT 2.45, ONNX Runtime 1.27.1 | [Download](https://china-qaihub-public-assets.s3.cn-northwest-1.amazonaws.com.cn/qai-hub-models/models/levit/releases/v0.62.1/levit-onnx-w8a16_mixed_int16.zip)
| ONNX | w8a8_mixed_int16 | Universal | QAIRT 2.45, ONNX Runtime 1.27.1 | [Download](https://china-qaihub-public-assets.s3.cn-northwest-1.amazonaws.com.cn/qai-hub-models/models/levit/releases/v0.62.1/levit-onnx-w8a8_mixed_int16.zip)
| QNN_DLC | float | Universal | QAIRT 2.45 | [Download](https://china-qaihub-public-assets.s3.cn-northwest-1.amazonaws.com.cn/qai-hub-models/models/levit/releases/v0.62.1/levit-qnn_dlc-float.zip)
| QNN_DLC | w8a16 | Universal | QAIRT 2.45 | [Download](https://china-qaihub-public-assets.s3.cn-northwest-1.amazonaws.com.cn/qai-hub-models/models/levit/releases/v0.62.1/levit-qnn_dlc-w8a16.zip)
| QNN_DLC | w8a16_mixed_int16 | Universal | QAIRT 2.45 | [Download](https://china-qaihub-public-assets.s3.cn-northwest-1.amazonaws.com.cn/qai-hub-models/models/levit/releases/v0.62.1/levit-qnn_dlc-w8a16_mixed_int16.zip)
| TFLITE | float | Universal | QAIRT 2.45 | [Download](https://china-qaihub-public-assets.s3.cn-northwest-1.amazonaws.com.cn/qai-hub-models/models/levit/releases/v0.62.1/levit-tflite-float.zip)

For more device-specific assets and performance metrics, visit **[LeViT on Qualcomm® AI Hub](https://aihub.qualcomm.cn/models/levit)**.


### Option 2: Export with Custom Configurations

Use the [Qualcomm® AI Hub Models](https://github.com/qualcomm/ai-hub-models-cn/blob/main/src/qai_hub_models/models/levit) Python library to compile and export the model with your own:
- Custom weights (e.g., fine-tuned checkpoints)
- Custom input shapes
- Target device and runtime configurations

This option is ideal if you need to customize the model beyond the default configuration provided here.

See our repository for [LeViT on GitHub](https://github.com/qualcomm/ai-hub-models-cn/blob/main/src/qai_hub_models/models/levit) for usage instructions.

## Model Details

**Model Type:** Model_use_case.image_classification

**Model Stats:**
- Input resolution: 224x224
- Model checkpoint: LeViT-128S
- Model size (float): 29.9 MB
- Model size (w8a16): 8.83 MB
- Number of parameters: 7.82M

## Performance Summary
| Model | Runtime | Precision | Chipset | Inference Time (ms) | Peak Memory Range (MB) | Primary Compute Unit
|---|---|---|---|---|---|---
| LeViT | ONNX | float | Snapdragon® X2 Elite | 0.7 ms | 2 - 2 MB | NPU
| LeViT | ONNX | float | Snapdragon® X Elite | 1.329 ms | 16 - 16 MB | NPU
| LeViT | ONNX | float | Snapdragon® 8 Gen 3 Mobile | 0.874 ms | 0 - 86 MB | NPU
| LeViT | ONNX | float | Snapdragon® 8 Gen 1 Mobile | 1.58 ms | 0 - 84 MB | NPU
| LeViT | ONNX | float | Qualcomm® Dragonwing™ IQ-8275 | 1.428 ms | 1 - 4 MB | NPU
| LeViT | ONNX | float | Qualcomm® Dragonwing™ QCS8550 (Proxy) | 1.28 ms | 0 - 25 MB | NPU
| LeViT | ONNX | float | Qualcomm® QCS8450 | 1.58 ms | 0 - 84 MB | NPU
| LeViT | ONNX | float | Qualcomm® Dragonwing™ IQ-9075 | 1.531 ms | 1 - 3 MB | NPU
| LeViT | ONNX | float | Qualcomm® Dragonwing™ IQ-X7181 | 1.329 ms | 16 - 16 MB | NPU
| LeViT | ONNX | float | Qualcomm® Dragonwing™ Q-8750 | 0.704 ms | 0 - 57 MB | NPU
| LeViT | ONNX | float | Snapdragon® 8 Elite Mobile | 0.704 ms | 0 - 57 MB | NPU
| LeViT | ONNX | float | Snapdragon® 8 Elite Gen 5 Mobile | 0.651 ms | 0 - 61 MB | NPU
| LeViT | ONNX | w8a16 | Snapdragon® X2 Elite | 0.487 ms | 1 - 1 MB | NPU
| LeViT | ONNX | w8a16 | Snapdragon® X Elite | 1.071 ms | 8 - 8 MB | NPU
| LeViT | ONNX | w8a16 | Snapdragon® 8 Gen 3 Mobile | 0.684 ms | 0 - 83 MB | NPU
| LeViT | ONNX | w8a16 | Snapdragon® 8 Gen 1 Mobile | 1.248 ms | 0 - 83 MB | NPU
| LeViT | ONNX | w8a16 | Qualcomm® Dragonwing™ QCS6490 | 4.427 ms | 0 - 3 MB | NPU
| LeViT | ONNX | w8a16 | Qualcomm® Dragonwing™ IQ-8275 | 1.025 ms | 0 - 4 MB | NPU
| LeViT | ONNX | w8a16 | Qualcomm® Dragonwing™ QCS8550 (Proxy) | 1.009 ms | 0 - 12 MB | NPU
| LeViT | ONNX | w8a16 | Qualcomm® QCS8450 | 1.248 ms | 0 - 83 MB | NPU
| LeViT | ONNX | w8a16 | Qualcomm® Dragonwing™ IQ-9075 | 1.171 ms | 0 - 3 MB | NPU
| LeViT | ONNX | w8a16 | Qualcomm® Dragonwing™ IQ-X7181 | 1.071 ms | 8 - 8 MB | NPU
| LeViT | ONNX | w8a16 | Qualcomm® Dragonwing™ Q-6690 | 4.31 ms | 0 - 178 MB | NPU
| LeViT | ONNX | w8a16 | Qualcomm® Dragonwing™ Q-7790 | 1.009 ms | 0 - 56 MB | NPU
| LeViT | ONNX | w8a16 | Qualcomm® Dragonwing™ Q-8750 | 0.517 ms | 0 - 56 MB | NPU
| LeViT | ONNX | w8a16 | Snapdragon® 8 Elite Mobile | 0.517 ms | 0 - 56 MB | NPU
| LeViT | ONNX | w8a16 | Snapdragon® 8 Elite Gen 5 Mobile | 0.47 ms | 0 - 58 MB | NPU
| LeViT | ONNX | w8a16 | Snapdragon® 7 Gen 4 Mobile | 1.009 ms | 0 - 56 MB | NPU
| LeViT | ONNX | w8a16_mixed_int16 | Snapdragon® X2 Elite | 0.509 ms | 1 - 1 MB | NPU
| LeViT | ONNX | w8a16_mixed_int16 | Snapdragon® X Elite | 1.114 ms | 10 - 10 MB | NPU
| LeViT | ONNX | w8a16_mixed_int16 | Snapdragon® 8 Gen 3 Mobile | 0.721 ms | 0 - 85 MB | NPU
| LeViT | ONNX | w8a16_mixed_int16 | Qualcomm® Dragonwing™ QCS8550 (Proxy) | 1.069 ms | 0 - 58 MB | NPU
| LeViT | ONNX | w8a16_mixed_int16 | Qualcomm® Dragonwing™ IQ-9075 | 1.217 ms | 0 - 3 MB | NPU
| LeViT | ONNX | w8a16_mixed_int16 | Qualcomm® Dragonwing™ IQ-X7181 | 1.114 ms | 10 - 10 MB | NPU
| LeViT | ONNX | w8a16_mixed_int16 | Qualcomm® Dragonwing™ Q-8750 | 0.549 ms | 0 - 55 MB | NPU
| LeViT | ONNX | w8a16_mixed_int16 | Snapdragon® 8 Elite Mobile | 0.549 ms | 0 - 55 MB | NPU
| LeViT | ONNX | w8a16_mixed_int16 | Snapdragon® 8 Elite Gen 5 Mobile | 0.497 ms | 0 - 63 MB | NPU
| LeViT | ONNX | w8a8_mixed_int16 | Snapdragon® 8 Gen 3 Mobile | 1.974 ms | 0 - 76 MB | NPU
| LeViT | ONNX | w8a8_mixed_int16 | Qualcomm® Dragonwing™ IQ-8275 | 3.924 ms | 7 - 10 MB | NPU
| LeViT | ONNX | w8a8_mixed_int16 | Qualcomm® Dragonwing™ QCS8550 (Proxy) | 2.626 ms | 0 - 30 MB | NPU
| LeViT | ONNX | w8a8_mixed_int16 | Qualcomm® Dragonwing™ IQ-9075 | 3.66 ms | 7 - 9 MB | NPU
| LeViT | ONNX | w8a8_mixed_int16 | Qualcomm® Dragonwing™ Q-8750 | 1.644 ms | 0 - 48 MB | NPU
| LeViT | ONNX | w8a8_mixed_int16 | Snapdragon® 8 Elite Mobile | 1.644 ms | 0 - 48 MB | NPU
| LeViT | ONNX | w8a8_mixed_int16 | Snapdragon® 8 Elite Gen 5 Mobile | 1.966 ms | 0 - 56 MB | NPU
| LeViT | QNN_DLC | float | Snapdragon® X2 Elite | 0.946 ms | 1 - 1 MB | NPU
| LeViT | QNN_DLC | float | Snapdragon® X Elite | 1.842 ms | 1 - 1 MB | NPU
| LeViT | QNN_DLC | float | Snapdragon® 8 Gen 3 Mobile | 1.084 ms | 0 - 85 MB | NPU
| LeViT | QNN_DLC | float | Snapdragon® 8 Gen 1 Mobile | 2.372 ms | 0 - 81 MB | NPU
| LeViT | QNN_DLC | float | Qualcomm® Dragonwing™ IQ-8275 | 1.661 ms | 1 - 4 MB | NPU
| LeViT | QNN_DLC | float | Qualcomm® Dragonwing™ QCS8550 (Proxy) | 1.596 ms | 0 - 2 MB | NPU
| LeViT | QNN_DLC | float | Qualcomm® SA8775P | 1.985 ms | 1 - 61 MB | NPU
| LeViT | QNN_DLC | float | Qualcomm® SA8650P | 1.985 ms | 1 - 61 MB | NPU
| LeViT | QNN_DLC | float | Qualcomm® SA8255P | 1.985 ms | 1 - 61 MB | NPU
| LeViT | QNN_DLC | float | Qualcomm® QCS8450 | 2.372 ms | 0 - 81 MB | NPU
| LeViT | QNN_DLC | float | Qualcomm® Dragonwing™ IQ-9075 | 1.819 ms | 1 - 3 MB | NPU
| LeViT | QNN_DLC | float | Qualcomm® Dragonwing™ IQ-X7181 | 1.842 ms | 1 - 1 MB | NPU
| LeViT | QNN_DLC | float | Qualcomm® Dragonwing™ Q-8750 | 0.831 ms | 0 - 58 MB | NPU
| LeViT | QNN_DLC | float | Qualcomm® SA7255P | 3.838 ms | 0 - 57 MB | NPU
| LeViT | QNN_DLC | float | Qualcomm® SA8295P | 2.349 ms | 0 - 52 MB | NPU
| LeViT | QNN_DLC | float | Snapdragon® 8 Elite Mobile | 0.831 ms | 0 - 58 MB | NPU
| LeViT | QNN_DLC | float | Snapdragon® 8 Elite Gen 5 Mobile | 0.72 ms | 1 - 61 MB | NPU
| LeViT | QNN_DLC | w8a16 | Snapdragon® X2 Elite | 0.783 ms | 0 - 0 MB | NPU
| LeViT | QNN_DLC | w8a16 | Snapdragon® X Elite | 1.587 ms | 0 - 0 MB | NPU
| LeViT | QNN_DLC | w8a16 | Snapdragon® 8 Gen 3 Mobile | 0.945 ms | 0 - 75 MB | NPU
| LeViT | QNN_DLC | w8a16 | Qualcomm® Dragonwing™ IQ-8275 | 1.343 ms | 0 - 3 MB | NPU
| LeViT | QNN_DLC | w8a16 | Qualcomm® Dragonwing™ QCS8550 (Proxy) | 1.389 ms | 0 - 29 MB | NPU
| LeViT | QNN_DLC | w8a16 | Qualcomm® SA8775P | 1.683 ms | 0 - 52 MB | NPU
| LeViT | QNN_DLC | w8a16 | Qualcomm® SA8650P | 1.683 ms | 0 - 52 MB | NPU
| LeViT | QNN_DLC | w8a16 | Qualcomm® SA8255P | 1.683 ms | 0 - 52 MB | NPU
| LeViT | QNN_DLC | w8a16 | Qualcomm® Dragonwing™ IQ-9075 | 1.529 ms | 0 - 2 MB | NPU
| LeViT | QNN_DLC | w8a16 | Qualcomm® Dragonwing™ IQ-X7181 | 1.587 ms | 0 - 0 MB | NPU
| LeViT | QNN_DLC | w8a16 | Qualcomm® Dragonwing™ Q-6690 | 5.542 ms | 0 - 174 MB | NPU
| LeViT | QNN_DLC | w8a16 | Qualcomm® Dragonwing™ Q-7790 | 1.415 ms | 0 - 50 MB | NPU
| LeViT | QNN_DLC | w8a16 | Qualcomm® Dragonwing™ Q-8750 | 0.66 ms | 0 - 47 MB | NPU
| LeViT | QNN_DLC | w8a16 | Qualcomm® SA7255P | 2.76 ms | 0 - 49 MB | NPU
| LeViT | QNN_DLC | w8a16 | Snapdragon® 8 Elite Mobile | 0.66 ms | 0 - 47 MB | NPU
| LeViT | QNN_DLC | w8a16 | Snapdragon® 8 Elite Gen 5 Mobile | 0.577 ms | 0 - 50 MB | NPU
| LeViT | QNN_DLC | w8a16 | Snapdragon® 7 Gen 4 Mobile | 1.415 ms | 0 - 50 MB | NPU
| LeViT | QNN_DLC | w8a16_mixed_int16 | Snapdragon® X2 Elite | 0.801 ms | 0 - 0 MB | NPU
| LeViT | QNN_DLC | w8a16_mixed_int16 | Snapdragon® X Elite | 1.652 ms | 0 - 0 MB | NPU
| LeViT | QNN_DLC | w8a16_mixed_int16 | Snapdragon® 8 Gen 3 Mobile | 0.973 ms | 0 - 75 MB | NPU
| LeViT | QNN_DLC | w8a16_mixed_int16 | Qualcomm® Dragonwing™ QCS8550 (Proxy) | 1.451 ms | 0 - 3 MB | NPU
| LeViT | QNN_DLC | w8a16_mixed_int16 | Qualcomm® SA8775P | 1.759 ms | 0 - 53 MB | NPU
| LeViT | QNN_DLC | w8a16_mixed_int16 | Qualcomm® SA8650P | 1.759 ms | 0 - 53 MB | NPU
| LeViT | QNN_DLC | w8a16_mixed_int16 | Qualcomm® SA8255P | 1.759 ms | 0 - 53 MB | NPU
| LeViT | QNN_DLC | w8a16_mixed_int16 | Qualcomm® Dragonwing™ IQ-9075 | 1.665 ms | 0 - 2 MB | NPU
| LeViT | QNN_DLC | w8a16_mixed_int16 | Qualcomm® Dragonwing™ IQ-X7181 | 1.652 ms | 0 - 0 MB | NPU
| LeViT | QNN_DLC | w8a16_mixed_int16 | Qualcomm® Dragonwing™ Q-6690 | 5.938 ms | 0 - 177 MB | NPU
| LeViT | QNN_DLC | w8a16_mixed_int16 | Qualcomm® Dragonwing™ Q-7790 | 1.435 ms | 0 - 51 MB | NPU
| LeViT | QNN_DLC | w8a16_mixed_int16 | Qualcomm® Dragonwing™ Q-8750 | 0.698 ms | 0 - 49 MB | NPU
| LeViT | QNN_DLC | w8a16_mixed_int16 | Qualcomm® SA7255P | 2.877 ms | 0 - 51 MB | NPU
| LeViT | QNN_DLC | w8a16_mixed_int16 | Snapdragon® 8 Elite Mobile | 0.698 ms | 0 - 49 MB | NPU
| LeViT | QNN_DLC | w8a16_mixed_int16 | Snapdragon® 8 Elite Gen 5 Mobile | 0.604 ms | 0 - 57 MB | NPU
| LeViT | QNN_DLC | w8a16_mixed_int16 | Snapdragon® 7 Gen 4 Mobile | 1.435 ms | 0 - 51 MB | NPU
| LeViT | TFLITE | float | Snapdragon® 8 Gen 3 Mobile | 1.037 ms | 0 - 92 MB | NPU
| LeViT | TFLITE | float | Snapdragon® 8 Gen 1 Mobile | 2.348 ms | 0 - 87 MB | NPU
| LeViT | TFLITE | float | Qualcomm® Dragonwing™ IQ-8275 | 1.674 ms | 0 - 20 MB | NPU
| LeViT | TFLITE | float | Qualcomm® Dragonwing™ QCS8550 (Proxy) | 1.511 ms | 0 - 3 MB | NPU
| LeViT | TFLITE | float | Qualcomm® SA8775P | 1.996 ms | 0 - 63 MB | NPU
| LeViT | TFLITE | float | Qualcomm® SA8650P | 1.996 ms | 0 - 63 MB | NPU
| LeViT | TFLITE | float | Qualcomm® SA8255P | 1.996 ms | 0 - 63 MB | NPU
| LeViT | TFLITE | float | Qualcomm® QCS8450 | 2.348 ms | 0 - 87 MB | NPU
| LeViT | TFLITE | float | Qualcomm® Dragonwing™ IQ-9075 | 1.81 ms | 0 - 19 MB | NPU
| LeViT | TFLITE | float | Qualcomm® Dragonwing™ Q-8750 | 0.781 ms | 0 - 70 MB | NPU
| LeViT | TFLITE | float | Qualcomm® SA7255P | 4.029 ms | 0 - 62 MB | NPU
| LeViT | TFLITE | float | Qualcomm® SA8295P | 2.341 ms | 0 - 53 MB | NPU
| LeViT | TFLITE | float | Snapdragon® 8 Elite Mobile | 0.781 ms | 0 - 70 MB | NPU
| LeViT | TFLITE | float | Snapdragon® 8 Elite Gen 5 Mobile | 0.665 ms | 0 - 71 MB | NPU

## License
* The license for the original implementation of LeViT can be found
  [here](https://github.com/facebookresearch/LeViT?tab=Apache-2.0-1-ov-file).

## References
* [LeViT: a Vision Transformer in ConvNet's Clothing for Faster Inference](https://arxiv.org/abs/2104.01136)
* [Source Model Implementation](https://github.com/facebookresearch/LeViT)

## Community
* For questions or feedback, chat with our engineers in [Qualcomm® AI Hub Workbench](https://workbench.aihub.qualcomm.cn).