# Qualcomm IM SDK overview The Qualcomm^®^ Intelligent Multimedia SDK (IM SDK) provides Qualcomm hardware-accelerated GStreamer-based plugins for optimized application development and includes reference applications that can be used to develop various use cases. Additionally, Qualcomm packages compatible versions of AI SDKs, namely, Qualcomm^®^ Neural Processing SDK, Qualcomm^®^ AI Engine direct SDK, and Lite Runtime or LiteRT (formerly known as TensorFlow Lite). Tab Qualcomm Linux Tab Ubuntu - *class* tabincludedirectiv - ## Architecture The figure shows the various software components of the Qualcomm IM SDK. **Figure : Qualcomm IM SDK software stack** > > > ![code](data:image/png;base64,UklGRogAAABXRUJQVlA4THwAAAAvF8AFEF8QEUrbNoJzhz90KneUDGn+4Ca2bafaZOieBCS8lsEEEuiRkHskIOHnHK69S3IQ0f8JYPvyL+ohhLfMVvB3UdMDCLOyk/WNP/dIrtmMkaCYHG0Y4ZPM1PCNNSNBIZ7GI7nmzFjPenP6Ln9XSw/wltkKyr92CNke) [QIM source code](https://git.codelinaro.org/clo/le/platform/vendor/qcom-opensource/gst-plugins-qti-oss/-/tree/imsdk.lnx.2.0.0.r2-rel?ref_type=heads) ## Qualcomm IM SDK components The components of the Qualcomm IM SDK are listed as follows: Components of Qualcomm IM SDK | Components | Description | | --- | --- | | GStreamer plugins | Provides hardware-accelerated GStreamer-based plugins for multimedia functions and image pre- and post-processing utilities required by the AI/ML models. | | Sample applications | Includes C++ and Python GStreamer applications for AI, and multimedia use cases. For more information, see [Sample applications](https://docs.qualcomm.com/bundle/publicresource/topics/80-70020-50/example-applications.html). | | Qualcomm Neural Processing SDK |

It is also known as Snapdragon Neural Processing Engine
(SNPE)

It is a software-accelerated runtime for executing deep
neural networks.

SNPE offers tools to convert and quantize neural
networks, and accelerate them on hardware accelerators
including CPU, GPU, and Qualcomm^® Hexagon^™ Tensor
Processor.

| | Qualcomm AI Engine direct SDK |

It provides a unified Qualcomm Neural Network (QNN) API
and tools to accelerate the AI/ML models, use cases on
Qualcomm chipsets and AI acceleration cores.

It provides a unified API and modular and extensible
per-accelerator libraries, which form a reusable basis
for full-stack AI solutions.

It supports runtimes such as Qualcomm Neural Processing
SDK and LiteRT AI Engine direct delegate.

| | LiteRT | LiteRT is an open-source deep learning framework
designed for on-device inference. Qualcomm IM SDK includes delegates
and tools from the compatible LiteRT package. | Note - See [Hardware SoCs](https://docs.qualcomm.com/bundle/publicresource/topics/80-70020-115/soc.html) that are supported on Qualcomm Linux. - For information about new features, sample applications, resolved issues, and limitations, see [Qualcomm Intelligent Multimedia SDK (IM SDK) Release Notes](https://docs.qualcomm.com/bundle/publicresource/topics/80-70020-52). - This guide isn't applicable for Qualcomm Dragonwing^TM^ IQ-615. ## Run sample applications The Qualcomm IM SDK sample applications show how to use the SDKs and develop end-to-end edge analytics use cases. For more information about the sample applications for Qualcomm Linux, see [Sample applications](https://docs.qualcomm.com/bundle/publicresource/topics/80-70020-50/example-applications.html). ## Platform layer The Qualcomm IM SDK supports Qualcomm Linux. For more information about Qualcomm Linux, see [System Software Overview User Guide](https://docs.qualcomm.com/bundle/publicresource/topics/80-70020-252/qualcomm-linux-sw-overview.html). ## Development kits Development kits are Qualcomm reference devices built to support Qualcomm Linux. They're used to develop and test software applications. A development kit includes necessary hardware and interfaces such as sensors, inertial measurement units (IMU), and other peripherals. You can use these development kits to test the capabilities of Qualcomm Linux. For more information, see [Development Kits](https://docs.qualcomm.com/bundle/publicresource/topics/80-70020-115/dev-kits.html). ## Supported component versions The current release of the Qualcomm IM SDK integrates the following components: | Components | Version | | --- | --- | | Qualcomm Neural Processing SDK | 2.35.0.250530 | | Qualcomm AI Engine direct SDK | 2.35.0.250530 | | LiteRT | 2.16.1 | | GStreamer | 1.22.12 | ## Qualcomm IM eSDK The eSDK is a package generated from the Qualcomm Linux image. It's installed on an Ubuntu host and provides a Yocto‑based environment that you can use for application development. The eSDK uses the [devtool](https://docs.yoctoproject.org/ref-manual/devtool-reference.html) mechanism of Yocto Project to download, compile, and install reference applications and plugins. The following figure shows the components of the Qualcomm IM eSDK. **Figure : Qualcomm IM eSDK** - *class* tabincludedirectiv - ## Architecture The figure shows the various software components of the Qualcomm IM SDK. **Figure : Qualcomm IM SDK software stack** ## Qualcomm IM SDK components The components of the Qualcomm IM SDK are listed as follows: Components of Qualcomm IM SDK | Components | Description | | --- | --- | | GStreamer plugins | Provides hardware-accelerated GStreamer-based plugins for multimedia functions and image pre- and post-processing utilities required by the AI/ML models. | | Sample applications | Includes C++ and Python GStreamer applications for AI, and multimedia use cases. For more information, see [Sample applications](https://docs.qualcomm.com/bundle/publicresource/topics/80-70020-50/example-applications.html). | | Qualcomm Neural Processing SDK |

It is also known as Snapdragon Neural Processing Engine
(SNPE)

It is a software-accelerated runtime for executing deep
neural networks.

SNPE offers tools to convert and quantize neural
networks, and accelerate them on hardware accelerators
including CPU, GPU, and Qualcomm^® Hexagon^™ Tensor
Processor.

| | Qualcomm AI Engine direct SDK |

It provides a unified Qualcomm Neural Network (QNN) API
and tools to accelerate the AI/ML models, use cases on
Qualcomm chipsets and AI acceleration cores.

It provides a unified API and modular and extensible
per-accelerator libraries, which form a reusable basis
for full-stack AI solutions.

It supports runtimes such as Qualcomm Neural Processing
SDK and LiteRT AI Engine direct delegate.

| | LiteRT | LiteRT is an open-source deep learning framework
designed for on-device inference. Qualcomm IM SDK includes delegates
and tools from the compatible LiteRT package. | Note The commands in this guide are compatible with [Ubuntu Server image](https://people.canonical.com/~platform/images/qualcomm-iot/ubuntu-24.04/ubuntu-24.04-x06/ubuntu-server-24.04/) for [Qualcomm Dragonwing™ RB3 Gen 2 Vision Development Kit](https://ubuntu.com/download/qualcomm-iot). ## Run sample applications The Qualcomm IM SDK sample applications show how to use the SDKs and develop end-to-end edge analytics use cases. For more information about the sample applications for Ubuntu, see [Sample applications](https://docs.qualcomm.com/bundle/publicresource/topics/80-70020-50/example-applications.html). ## Platform layer Qualcomm IM SDK is supported on the certified Ubuntu image provided by Canonical. For more information about certified Canonical images, see [Ubuntu Certified devices](https://ubuntu.com/certified). ## Supported component versions The current release of the Qualcomm IM SDK integrates the following components: | Components | Personal Package Archive (PPA) | Version | | --- | --- | --- | | Qualcomm Neural Processing SDK | libsnpe1 | 2.39.0.250926 | | Qualcomm AI Engine direct SDK | libqnn1 | 2.39.0.250926 | | LiteRT | libtensorflow-lite-c-qcom1 | 2.16.1 | | GStreamer | libgstreamer-qcom1 | 1.24.2 | Last Published: Jan 18, 2026 [Previous Topic Qualcomm IM SDK quickstart documentation](https://docs.qualcomm.com/bundle/publicresource/80-70020-51/topics/qmi-sdk-qsg-landing-page.md) [Next Topic Get started with the Qualcomm IM SDK](https://docs.qualcomm.com/bundle/publicresource/80-70020-51/topics/install-sdk.md)