Qualcomm AI Engine Direct
QualcommQualcomm AI Engine Direct, also called the Qualcomm AI Runtime (QAIRT or QNN), is an on-device inference SDK for Snapdragon processors, distributed as a separate SDK download from Qualcomm. It exposes a unified low-level API over the Hexagon NPU/HTP, Adreno GPU, and Kryo CPU. LiteRT, ONNX Runtime, and ExecuTorch delegate to it as their Qualcomm NPU backend. Qualcomm's qai-appbuilder, formerly ai-engine-direct-helper, wraps it in a convenience API on GitHub.
Qualcomm's own docs and Software Center pages render as JavaScript shells and could not be read directly, so the SDK's behavior was confirmed from the integrating projects' documentation. Verified 2026-08-09 via GitHub, the ExecuTorch Qualcomm backend docs, and the LiteRT NPU documentation.
Openness
1 high confidence- source
- closed(no published source for the QNN/Qualcomm AI Runtime SDK core engine
- license
- proprietary(no license to the source since it is not published)
- open-wrapper
- BSD-3-Clause(qualcomm/qai-appbuilder, formerly quic/ai-engine-direct-helper -- a convenience API layer that requires the proprietary SDK libraries)
The core engine is closed: there is no published repository for the QNN / Qualcomm AI Runtime SDK, which ships only as an EULA-gated binary download from the Qualcomm Software Center, and no license is granted to source that is not published. There is an open wrapper, qualcomm/qai-appbuilder, formerly quic/ai-engine-direct-helper, with both names now redirecting to the same place. Under its current name that wrapper is BSD-3-Clause, but it states that "Some libraries from the Qualcomm AI Runtime SDK are required to use QAI AppBuilder", so the open piece is a convenience API layer whose dependency is still an EULA-gated binary. That does not change the class of the engine itself.
- https://github.com/qualcomm/qai-appbuilder recorded 2026-08-13
The wrapper under its current name: BSD 3-Clause, and "Some libraries from the Qualcomm® AI Runtime SDK are required to use QAI AppBuilder" - an open wrapper over a proprietary SDK that must be downloaded separately from Qualcomm's Software Center.
- https://github.com/quic/ai-engine-direct-helper recorded 2026-08-09
the former name, kept because the openness note turns on the move; redirects to qualcomm/qai-appbuilder
Adoption
4 low confidenceIntegration breadth across the Snapdragon ecosystem: LiteRT (Google AI Edge) documents Snapdragon 8 Gen 1/2/3 and 8 Elite support, ExecuTorch delegates to it as the Qualcomm NPU backend, and ONNX Runtime ships a QNN execution provider. No developer, device, or download figure is published, so no numeric reach is recorded and the band rests on shipping-integration breadth rather than a reported figure.
- https://docs.pytorch.org/executorch/stable/backends-qualcomm.html recorded 2026-08-09
Currently, this ExecuTorch Backend can delegate AI computations to Hexagon processors and Adreno GPU through Qualcomm AI Engine Direct APIs.
- https://ai.google.dev/edge/litert/android/npu/qualcomm recorded 2026-08-09
essential to running inference on the NPU for your LiteRT model on-device. Supported devices include: Snapdragon 8 Gen 1 (SM8450) Snapdragon 8 Gen 2 (SM8550) Snapdragon 8 Gen 3 (SM8650) Snapdragon 8 Elite (SM8750) and more
- https://onnxruntime.ai/docs/execution-providers/QNN-ExecutionProvider.html recorded 2026-08-09
The QNN Execution Provider for ONNX Runtime enables hardware accelerated execution on Qualcomm chipsets. It uses the Qualcomm AI Engine Direct SDK (QNN SDK) to construct a QNN graph from an ONNX model
Capability
4 medium confidenceA leading mobile and edge on-device runtime: a unified low-level API across the Hexagon HTP/NPU, the Adreno GPU and the Kryo CPU, ahead-of-time and on-device compilation, and quantized INT8/INT16 LLM and VLM inference such as Gemma and FastVLM, with NPU speedups reported up to 100x over CPU and 10x over GPU on Snapdragon 8 Elite Gen 5. LiteRT, ExecuTorch and ONNX Runtime all delegate to it as their Qualcomm NPU backend. It remains Snapdragon-locked and edge-scale, which holds it below the server frontier engines in this category.
- https://docs.pytorch.org/executorch/stable/backends-qualcomm.html recorded 2026-08-09
Qualcomm AI Engine Direct is designed to provide unified, low-level APIs for AI development. Developers can interact with various accelerators on Qualcomm SoCs with these set of APIs, including Kryo CPU, Adreno GPU, and Hexagon processors.
- https://developers.googleblog.com/unlocking-peak-performance-on-qualcomm-npu-with-litert/ recorded 2026-08-09
You can now deploy your model seamlessly across all supported devices, with either ahead-of-time (AOT) or on-device compilation. ... the NPU acceleration provides up to a 100x speedup over CPU and a 10x speedup over GPU ... achieving SOTA performance for models like Gemma and FastVLM
Verified 2026-08-09