# GStreamer ML plugin and AI SDK APIs
Qualcomm Linux provides [Qualcomm® Intelligent Multimedia SDK (IM SDK) GStreamer plugin APIs](https://docs.qualcomm.com/bundle/publicresource/topics/80-70022-50/qim-sdk-plugins.html)
to interface with Qualcomm AI Runtime SDK (QAIRT) APIs, to optimally run deep learning models on hardware modules such as neural Processing Unit (NPU),
graphics Processing Unit (GPU), and central Processing Unit (CPU).
The following software block diagram shows how AI applications can call Qualcomm IM SDK GStreamer Plugin APIs (`qtimlsnpe`, `qtimlqnn`, `qtimltflite`) to interface with and access
QAIRT APIs to use a specific hardware accelerator.
## GStreamer Plugin APIs
Three GStreamer ML plugin APIs support the QAIRT SDK and TFLite, which provides flexibility for you to choose the right combination for your needs.
| [Qualcomm IM SDK plugin for Qualcomm Neural Processing Engine (qtimlsnpe)](https://docs.qualcomm.com/bundle/publicresource/topics/80-70022-50/qtimlsnpe.html) | Uses [Qualcomm Neural Processing Engine](https://docs.qualcomm.com/bundle/publicresource/topics/80-63442-2) APIs to load and execute models.
Choose this plugin for quick prototyping and high-level API support. |
| --- | --- |
| [Qualcomm IM SDK plugin for Qualcomm AI Engine Direct plugin (qtimlqnn)](https://docs.qualcomm.com/bundle/publicresource/topics/80-70022-50/qtimlqnn.html) | Uses [Qualcomm AI Engine Direct](https://docs.qualcomm.com/bundle/publicresource/topics/80-63442-50) APIs, which provide low-level, unified API and improved performance to optimize and execute network models on the desired hardware accelerator.
Choose this plugin for advanced graph execution options and optimizations. |
| [Qualcomm IM SDK plugin for LiteRT (qtimltflite)](https://docs.qualcomm.com/bundle/publicresource/topics/80-70022-50/qtimltflite.html) | Accelerates LiteRT models directly using [Qualcomm AI Engine Direct](https://docs.qualcomm.com/bundle/publicresource/topics/80-63442-50) APIs to load and execute models.
Choose this plugin to directly run LiteRT models, without conversion. |
## qualcomm AI Runtime SDK (QAIRT) APIs
Use these APIs to develop AI applications.
| [Qualcomm Neural Processing Engine SDK](https://docs.qualcomm.com/bundle/publicresource/topics/80-63442-2/api.html) | C, C++, and Java APIs to support heterogenous computing, system-level configurations, and direct AI workloads to all accelerator cores.
Provides you with flexibility, such as inter-core collaboration support. |
| --- | --- |
| [Qualcomm AI Engine Direct](https://docs.qualcomm.com/bundle/publicresource/topics/80-63442-50/api.html) | Low-level, highly customizable APIs that speed up AI models on all AI accelerator cores with individual libraries.
Use these APIs to target a specific accelerator core or delegate workloads from popular runtimes including the Qualcomm Neural Processing Engine SDK, LiteRT, and ONNX runtimes. |
Last Published: Sep 15, 2025
[Previous Topic
AI software overview and architecture](https://docs.qualcomm.com/bundle/publicresource/80-70022-15/topics/overview.md) [Next Topic
Hardware-accelerated sample applications for AI](https://docs.qualcomm.com/bundle/publicresource/80-70022-15/topics/sample-apps.md)