# QNN GPU Op Support Revision History

| Introduced in QNN SDK Version | Description |
| --- | --- |
| 2.39.0 | <ul class="simple"><br><li><p>Support INT_32 inputs/outputs in StridedSlice Operation.</p></li><br></ul> |
| 2.38.0 | <ul class="simple"><br><li><p>Support GatherNd Operation.</p></li><br><li><p>Support INT_32 datatype for Argmax and Argmin operation</p></li><br><li><p>Support UINT_32 datatype for Argmax and Argmin operation</p></li><br><li><p>Support CumulativeSum Operation.</p></li><br></ul> |
| 2.34.0 | <ul class="simple"><br><li><p>Support GridSample Operation for input and output rank = 4</p></li><br></ul> |
| 2.32.0 | <ul class="simple"><br><li><p>Support ElementWiseSoftplus and ElementWiseNeuron with SOFTPLUS parameter.</p></li><br></ul> |
| 2.28.0 | <ul class="simple"><br><li><p>Support Resize Operation for Nearest and Linear interpolation modes.</p></li><br><li><p>Support HardSigmoid Activation.</p></li><br></ul> |
| 2.26.0 | <ul class="simple"><br><li><p>Support ElementWiseSign Operation and ElementWiseUnary Operation with SIGN parameter</p></li><br><li><p>Support ScatterElements Operation.</p></li><br></ul> |
| 2.25.0 | <ul class="simple"><br><li><p>Support RmsNorm Operation for input rank &lt;= 4</p></li><br></ul> |
| 2.23.0 | <ul class="simple"><br><li><p>Support ElementWiseFloorDiv Operation and ElementwiseBinary Operation with FloorDiv</p></li><br></ul> |
| 2.22.0 | <ul class="simple"><br><li><p>Support QNN_DATATYPE_UFIXED_POINT_4 in ElementwiseSelect op.</p></li><br><li><p>Support QNN_DATATYPE_UFIXED_POINT_4 in BinaryElementwise ops.</p></li><br><li><p>Support Concat op with input rank 5 and axis = 1 if batch = 1</p></li><br><li><p>Support QNN_DATATYPE_UFIXED_POINT_4 in Gather op.</p></li><br></ul> |
| 2.21.0 | <ul class="simple"><br><li><p>Support QNN_DATATYPE_BOOL_8 inputs/outputs to Reshape op.</p></li><br><li><p>Support QNN_DATATYPE_BOOL_8 outputs to Cast op.</p></li><br><li><p>Support QNN_DATATYPE_INT_32 in Concat op.</p></li><br><li><p>Support QNN_DATATYPE_INT_32 in BinaryElementwise op.</p></li><br><li><p>Support QNN_DATATYPE_UFIXED_POINT_4 in Dequantize op.</p></li><br><li><p>Support ScatterND Operation.</p></li><br></ul> |
| 2.20.0 | <ul class="simple"><br><li><p>Support GroupNorm with input rank = {3,4}</p></li><br></ul> |
| 2.18.0 | <ul class="simple"><br><li><p>Support LayerNorm non-depth axis normalization and batch != 1</p></li><br></ul> |
| 2.16.0 | <ul class="simple"><br><li><p>Support 5D inputs in StridedSlice operation.</p></li><br><li><p>Support Concat op for input rank = 5 and axis &gt; 1.</p></li><br><li><p>Support ElementWiseNeuron Operation</p></li><br></ul> |
| 2.14.0 | <ul class="simple"><br><li><p>Support BOOL_8 inputs to Cast operation.</p></li><br><li><p>Support CRD Mode in DepthToSpace Operation.</p></li><br><li><p>Support ElementWiseUnary Operation</p></li><br><li><p>Support ElementWiseBinary Operation</p></li><br></ul> |
| 2.13.0 | <ul class="simple"><br><li><p>Support LayerNorm for input rank &lt; 4 and batch &gt; 1 for rank 4 input.</p></li><br><li><p>Support Batch Broadcasting in MatMul Operation.</p></li><br><li><p>Support QNN_DATATYPE_INT_64 input datatype in Cast Operation.</p></li><br><li><p>Support reduction across batch</p></li><br><li><p>Support Broadcasting for ElementwiseSelect Operation.</p></li><br></ul> |
| 2.12.0 | <ul class="simple"><br><li><p>Support InstanceNorm Operation with 3D inputs</p></li><br><li><p>Support GELU Operation</p></li><br></ul> |
| 2.10.0 | <ul class="simple"><br><li><p>Support Pack operation with 1 input.</p></li><br></ul> |
| 2.9.0 | <ul class="simple"><br><li><p>Support FLOAT_16 inputs/outputs in Lstm Operation</p></li><br><li><p>Support non multiple of 4 inputs size in Lstm Operation</p></li><br></ul> |
| 2.3.0 | <ul class="simple"><br><li><p>Support INT_8 inputs/outputs in Cast Operation.</p></li><br><li><p>Support UINT_32 inputs/outputs in StridedSlice Operation.</p></li><br><li><p>Support INT_32 inputs/outputs in Transpose Operation.</p></li><br><li><p>Support UINT_8 inputs/outputs in Reshape Operation.</p></li><br><li><p>Update Tile operation validation to disallow tiling across batch dimension.</p></li><br></ul> |
| 2.0.0 | <ul class="simple"><br><li><p>Branched from QNN SDK version 1.15.0</p></li><br></ul> |

Last Published: Oct 10, 2025

[Previous Topic
QNN DSP Op Definition Revision History](https://docs.qualcomm.com/bundle/publicresource/80-63442-50/topics/dsp_opdef_version_history.md) [Next Topic
QNN HTA Op Definition Revision History](https://docs.qualcomm.com/bundle/publicresource/80-63442-50/topics/hta_opdef_version_history.md)