# QNN LPAI API Revision History

| QNN LPAI API Version | Introduced in QNN SDK Version | Model Builder Version | Description |
| --- | --- | --- | --- |
| 2.22.0 | 2.39.0 | 5.12.0 | <ul class="simple"><br><li><p>Add core selection config option</p></li><br><li><p>Support multi-type attributes</p></li><br><li><p>enable dspqueue for enpuV6 and enpuV5_1</p></li><br></ul> |
| 2.21.0 | 2.38.0 | 5.11.2d1a28fc | <ul class="simple"><br><li><p>Profiling improvements</p></li><br><li><p>PSRAM memory support</p></li><br><li><p>Support Async Execution</p></li><br></ul> |
| 2.20.0 | 2.37.0 | 5.10.8421ef45 | <ul class="simple"><br><li><p>Add ubuf config option</p></li><br><li><p>Improvements on profiling</p></li><br></ul> |
| 2.19.0 | 2.36.0 | 5.9.0c847578 | <ul class="simple"><br><li><p>LPAI BE adapted to work with 5.9 enpu components</p></li><br><li><p>More profiling improvements</p></li><br><li><p>Reset Call  handling improvements</p></li><br><li><p>Added OP support: RmsNorm, Buffer Op Padding Param</p></li><br></ul> |
| 2.18.0 | 2.35.0 | 5.8.35a52877 | <ul class="simple"><br><li><p>LPAI BE adapted to work with 5.8 enpu components</p></li><br><li><p>More profiling improvements</p></li><br><li><p>Added OP support: StatelessLSTM/GRU</p></li><br></ul> |
| 2.17.0 | 2.34.0 | 5.7.9931d1e5 | <ul class="simple"><br><li><p>Simplified Lpai BE library names</p></li><br><li><p>LPAI Documentation improvements</p></li><br><li><p>enpu compile,execute, simulator libraries update</p></li><br><li><p>LPAI BE dapted to work with 5.7 enpu components update</p></li><br><li><p>More profiling improvements</p></li><br><li><p>Added OP support:BatchToSpace,SpaceToBatch,DepthToSpace,SpaceToDepth,Quantize,Dequantize</p></li><br></ul> |
| 2.16.0 | 2.32.0 | 5.6.057f0428 | <ul class="simple"><br><li><p>Add device discovery mechanism across different LPAI Supported HW</p></li><br><li><p>Add support for LPAI graph early termination</p></li><br><li><p>Fix runtime stability issues</p></li><br><li><p>Improved Profiling information</p></li><br><li><p>Added OP support: Channel_Shuffle</p></li><br></ul> |
| 2.15.0 | 2.31.0 | 5.5.addf3789 | <ul class="simple"><br><li><p>Add Compiler + Sim for x86 Windows</p></li><br><li><p>Fix fastRPC bugs for hex79,hex81</p></li><br><li><p>Documentation</p></li><br><li><p>mem Register/deRegister</p></li><br></ul> |
| 2.14.0 | 2.30.0 | 5.4.ad7bd0ed | <ul class="simple"><br><li><p>Add Support for Backend Custom properties</p></li><br></ul> |
| 2.13.1 | 2.29.0 | 4.14.f7310daf | <ul class="simple"><br><li><p>Add QNN LPAI island interface</p></li><br><li><p>Add model preparation and simulation support on x86 Windows platform</p></li><br></ul> |
| 2.13.0 | 2.28.0 | 4.13.efc48af1 | <ul class="simple"><br><li><p>Align QNN LPAI config file defaults</p></li><br><li><p>Runtime check for compiled LPAI models older than v4.6, no longer supported</p></li><br><li><p>Prepare for LPAI Runtime Island Support</p></li><br></ul> |
| 2.12.0 | 2.27.0 | 4.12.5bf898df | <ul class="simple"><br><li><p>Add profiling support: graph execute time</p></li><br><li><p>Bug fix for unsupported float tensor</p></li><br><li><p>Bug fix for reset tensors as input</p></li><br></ul> |
| 2.11.0 | 2.26.1 | 4.11.7fa40f05 | <ul class="simple"><br><li><p>Op update: Add default support of antialias param for ResizeBilinear Op</p></li><br></ul> |
| 2.11.0 | 2.25.1 | 4.11.7fa40f05 | <ul class="simple"><br><li><p>Ops updates and bug-fixes: Split,LSTM,GRU,Batchnorm,Conv,StridedSlice,ReduceProd</p></li><br><li><p>Add profiling support: per layer execution time, layer fusion info, layer linking info</p></li><br></ul> |
| 2.10.0 | 2.24.1 | 4.10.68490b7c | <ul class="simple"><br><li><p>Add Op support: Buffer (Framer)</p></li><br><li><p>PerLayer Profiling support via configs</p></li><br><li><p>Version Checks for LPAI Execute</p></li><br><li><p>Start support for lightweight internal aDSP build</p></li><br><li><p>Memory allocation for internal aDSP Execute</p></li><br></ul> |
| 2.9.0 | 2.23.1 | 4.9.0xa3f3a016 | <ul class="simple"><br><li><p>Optimize memory planning</p></li><br><li><p>Fix memory leaks issue during model conversion</p></li><br><li><p>Add Op support: Power, Layernorm, Split</p></li><br><li><p>Fixed memory leak 32 properly if initialization is failed</p></li><br><li><p>Fixed stability issue related to multiple threads/processes</p></li><br><li><p>Add API profiler parameter</p></li><br></ul> |
| 2.8.0 | 2.22.1 | 4.8.42ed4bf0-v79 | <ul class="simple"><br><li><p>Added: Runtime eNPU5 Support, x86 generation, x86 Simulator and ARM execution</p></li><br></ul> |
| 2.7.1 | 2.20.1 | 3.17.4446b3115 | <ul class="simple"><br><li><p>Added: LPAI BE Adaptor getMap() APIs</p></li><br></ul> |
| 2.7.0 | 2.18.1 | 3.17.4446b311 | <ul class="simple"><br><li><p>Added: Support to model visualizer, eNPU4 for linux_x86, windows,hexagonsim and adsp</p></li><br><li><p>Added: LPAI BE getSupportedOperations()</p></li><br><li><p>Fixes: TensorID duplicate errors during QNN change from tensor name to tensorID</p></li><br><li><p>Fixes: Sanitize quantizations with invalid parameters</p></li><br></ul> |
| 2.6.0 | 2.17.1 | 3.16.ae82aa87 | <ul class="simple"><br><li><p>Fixes: Fix resize op to account for annotations on sizes tensor</p></li><br><li><p>Added: Support to multiple loaded HW compiler versions</p></li><br></ul> |
| 2.5.0 | 2.14.1 | 3.14.827f8f95 | <ul class="simple"><br><li><p>Fixes: Fix resize op to account for annotations on sizes tensor</p></li><br><li><p>Fixes: Deconv2D segfault data size calculated from model builder is incorrect.</p></li><br><li><p>Fixes: Fix segfault in case of min/max with constant tensor.</p></li><br><li><p>Fixes: Expand tensor shape when rank is different.</p></li><br></ul> |
| 2.4.0 | 2.13.1 | 3.12.a4c0bd91 | <ul class="simple"><br><li><p>Fixes: Proper bias alignment for gemm with square output</p></li><br><li><p>Fixes: Improve layernorm accuracy for small hidden_size</p></li><br><li><p>Added: Support to different permutations of 1d bias matmul.</p></li><br><li><p>Added: New memory planning compiler pipeline</p></li><br></ul> |
| 2.3.0 | 2.12.1 | 3.10.e4d3cfa4 | <ul class="simple"><br><li><p>Fixes: per-channel quantization</p></li><br></ul> |
| 2.2.0 | 2.11.1 | 3.8.74ea5d84 | <ul class="simple"><br><li><p>See op support updates</p></li><br></ul> |
| 2.1.0 | 2.10.1 | 3.6.546129c7 | <ul class="simple"><br><li><p>Fixes: symmetric quantization and compression ratio calc</p></li><br><li><p>Added: support multiple nodes with same id, input name from onnx node for perlayer dump</p></li><br></ul> |
| 2.0.0 | 2.9.1 |  | <ul class="simple"><br><li><p>Initial release</p></li><br></ul> |

Last Published: Oct 10, 2025

[Previous Topic
HTP Version History](https://docs.qualcomm.com/bundle/publicresource/80-63442-50/topics/htp_api_version_history.md) [Next Topic
QNN System API Revision History](https://docs.qualcomm.com/bundle/publicresource/80-63442-50/topics/api_version_history.md)