# Release Notes

This page contains the release notes for Qualcomm® AI Engine Direct Delegate:

> 
> 
> | **Version** | **Release Date** | **Key Updates** |
> | --- | --- | --- |
> | v2.39.0 | 09/30/2025 | <ul class="simple"><br><li><p>Support new SoC</p></li><br><li><p>Minor fixes</p></li><br></ul> |
> | v2.38.0 | 08/31/2025 | <ul class="simple"><br><li><p>Support new SoC</p></li><br><li><p>Add new java interface option: HTP_OPTIMIZE_FOR_INFERENCE_O3</p></li><br><li><p>Support Broadcast_to Op</p></li><br><li><p>Minor fixes</p></li><br></ul> |
> | v2.37.0 | 07/31/2025 | <ul class="simple"><br><li><p>Support new SoC</p></li><br><li><p>Minor fixes</p></li><br></ul> |
> | v2.36.0 | 06/30/2025 | <ul class="simple"><br><li><p>Support new SoC</p></li><br><li><p>Minor fixes</p></li><br></ul> |
> | v2.35.0 | 05/31/2025 | <ul class="simple"><br><li><p>Support new SoC</p></li><br><li><p>Minor fixes</p></li><br></ul> |
> | v2.34.0 | 04/30/2025 | <ul class="simple"><br><li><p>Support new SoC</p></li><br><li><p>Minor fixes</p></li><br></ul> |
> | v2.33.0 | 03/31/2025 | <ul class="simple"><br><li><p>Support new SoC</p></li><br><li><p>Minor fixes</p></li><br></ul> |
> | v2.32.0 | 02/28/2025 | <ul class="simple"><br><li><p>Support new SoC</p></li><br><li><p>Minor fixes</p></li><br></ul> |
> | v2.31.0 | 01/31/2025 | <ul class="simple"><br><li><p>Support IrBackend</p></li><br><li><p>Support EmbeddingLookup op</p></li><br></ul> |
> | v2.30.0 | 12/31/2024 | <ul class="simple"><br><li><p>Libraries for Android are compiled to be 16kb page-aligned</p></li><br><li><p>Support profiler in java interface</p></li><br><li><p>Support QNN context binary version 3</p></li><br><li><p><strong>NOTE</strong>: The default HTP precision is changed to HtpFp16</p></li><br><li><p><strong>NOTE</strong>: some chips don’t support HTP FP16 runtime</p></li><br></ul> |
> | v2.29.0 | 11/30/2024 | <ul class="simple"><br><li><p>Minor fixes</p></li><br></ul> |
> | v2.28.0 | 10/31/2024 | <ul class="simple"><br><li><p>TFLite renamed to LiteRT (TBD)</p></li><br><li><p>Revamped QNN Delegate logging</p></li><br><li><p>Added optimization strategy O3 to java interface</p></li><br><li><p>Mentioned Maven package and Proguard settings</p></li><br></ul> |
> | v2.27.0 | 09/30/2024 | <ul class="simple"><br><li><p>Added $ORIGIN to rpath</p></li><br><li><p>Added ‘packagin=aar’ in Android aar pom file</p></li><br><li><p>Fixed bug with HTP device_id affecting cache SHA</p></li><br></ul> |
> | v2.26.0 | 08/31/2024 | <ul class="simple"><br><li><p>Added O3 optimization option</p></li><br><li><p>Added linting profiling option</p></li><br><li><p>Added device ID setting</p></li><br></ul> |
> | v2.25.0 | 07/31/2024 | <ul class="simple"><br><li><p>Finished op-version check</p></li><br><li><p>Optimized power mode settings</p></li><br><li><p>Added license files into AAR</p></li><br></ul> |
> | v2.24.0 | 06/30/2024 | <ul class="simple"><br><li><p>OP version checks</p></li><br><li><p>Support TransposeConv2d with fusion</p></li><br><li><p>JNI statically linked</p></li><br><li><p>Support StridedSlice offset</p></li><br><li><p>Removed dynamic symbols</p></li><br><li><p>Plan to implement TFLite Stable Delegate APIs</p></li><br></ul> |
> | v2.23.0 | 05/31/2024 | <ul class="simple"><br><li><p>TF 2.15.0 and NDK26c used</p></li><br><li><p>Fixed supported operators docs</p></li><br><li><p>Optimized Square operator</p></li><br></ul> |
> | v2.22.0 | 04/30/2024 | <ul class="simple"><br><li><p>Added –htp_disable_conv_hmx option</p></li><br><li><p>Fixed QnnSaver bug</p></li><br><li><p>Support 16bit QuantizeOp</p></li><br><li><p>Support Square op</p></li><br><li><p>Improved capability detection</p></li><br><li><p>Plan to upgrade TF and NDK</p></li><br></ul> |
> | v2.21.0 | 03/31/2024 | <ul class="simple"><br><li><p>useConvHmx default True</p></li><br><li><p>TfLiteQnnDelegateDelete() not thread-safe</p></li><br></ul> |
> | v2.20.0 | 02/29/2024 | <ul class="simple"><br><li><p>Bug fixes around HTP Prepare</p></li><br><li><p>useConvHmx to be default True</p></li><br><li><p>TF upgrade planned</p></li><br></ul> |
> | v2.19.0 | 01/31/2024 | <ul class="simple"><br><li><p>Fixed RPC polling</p></li><br><li><p>Fixed skip nodes warnings</p></li><br></ul> |
> | v2.18.0 | 12/31/2023 | <ul class="simple"><br><li><p>Added java API checkCapability()</p></li><br><li><p>Compiled with NDK r25c</p></li><br><li><p>Added HTP vtcm size and hvx thread options</p></li><br></ul> |
> | v2.17.0 | 11/30/2023 | <ul class="simple"><br><li><p>Added graph priority</p></li><br><li><p>Snapdragon 8 Gen 3 support</p></li><br><li><p>Added java interface options</p></li><br><li><p>Improved fp16 fallback</p></li><br></ul> |
> | v2.16.0 | 10/06/2023 | <ul class="simple"><br><li><p>Added Atan2, AddN, Ceil</p></li><br><li><p>Toolchain change to NDK25c planned</p></li><br></ul> |
> | v2.15.0 | 09/30/2023 | <ul class="simple"><br><li><p>Added Conv3d, GatherNd, Elu, L2Pool2D, ScatternNd, Relu_0_to_1, TransposeConv3d, PadV2, OneHot, Cos, Round</p></li><br><li><p>Added Skip OP API example</p></li><br></ul> |
> | v2.14.0 | 08/31/2023 | <ul class="simple"><br><li><p>Added qtld-release.aar</p></li><br><li><p>Added platform support for aarch64-oe-linux-gcc11.2 and aarch64-ubuntu-gcc9.4</p></li><br></ul> |
> | v2.13.0 | 07/31/2023 | <ul class="simple"><br><li><p>Fixed preparation errors</p></li><br><li><p>Supported implicit group Conv2D</p></li><br></ul> |
> | v2.12.0 | 06/30/2023 | <ul class="simple"><br><li><p>Initial release</p></li><br><li><p>Built with ndk19c</p></li><br><li><p>Plan to switch to ndk25c</p></li><br></ul> |

Last Published: Oct 10, 2025

[Previous Topic
Qualcomm® AI Engine Direct Backend Library](https://docs.qualcomm.com/bundle/publicresource/80-63442-50/topics/qnn_libs.md) [Next Topic
Revision History](https://docs.qualcomm.com/bundle/publicresource/80-63442-50/topics/revision_history.md)