diff --git a/backends/nxp/runtime/NeutronBackend.cpp b/backends/nxp/runtime/NeutronBackend.cpp
index 3ea973b7c5b..02c9e0003be 100644
--- a/backends/nxp/runtime/NeutronBackend.cpp
+++ b/backends/nxp/runtime/NeutronBackend.cpp
@@ -95,6 +95,21 @@ void wait(uint32_t channel) {}
 static NeutronConfig neutronMemCopyConfig = {copy, wait};
 #endif
 
+// Extended dim_order check that handles 3D tensors (NCW → NWC).
+// The core is_channels_last_dim_order() only supports 4D/5D.
+bool is_channels_last_dim_order_extended(
+    const uint8_t* dim_order,
+    size_t dims) {
+  if (is_channels_last_dim_order(dim_order, dims)) {
+    return true;
+  }
+  // 3D: dim_order [0, 2, 1] means channel dim (1) is innermost → NWC layout.
+  if (dims == 3) {
+    return dim_order[0] == 0 && dim_order[1] == 2 && dim_order[2] == 1;
+  }
+  return false;
+}
+
 // Applied on outputs.
 template <typename T>
 void transposeToChannelFirst(
@@ -149,12 +164,21 @@ void transposeInput(
   if (length < 3) {
     return;
   }
-  size_t N = 1;
-  size_t C = sizes[length - 3];
-  size_t H = sizes[length - 2];
-  size_t W = sizes[length - 1];
-  for (size_t i = 0; i < length - 3; i++) {
-    N *= sizes[i];
+  size_t N, C, H, W;
+  if (length == 3) {
+    // 3D [N, C, W] → treat as [N, C, 1, W]
+    N = sizes[0];
+    C = sizes[1];
+    H = 1;
+    W = sizes[2];
+  } else {
+    N = 1;
+    C = sizes[length - 3];
+    H = sizes[length - 2];
+    W = sizes[length - 1];
+    for (size_t i = 0; i < length - 3; i++) {
+      N *= sizes[i];
+    }
   }
   switch (element_size) {
     case 1:
@@ -204,12 +228,21 @@ void transposeOutput(
   if (length < 3) {
     return;
   }
-  size_t N = 1;
-  size_t C = sizes[length - 3];
-  size_t H = sizes[length - 2];
-  size_t W = sizes[length - 1];
-  for (size_t i = 0; i < length - 3; i++) {
-    N *= sizes[i];
+  size_t N, C, H, W;
+  if (length == 3) {
+    // 3D [N, C, W] → treat as [N, C, 1, W]
+    N = sizes[0];
+    C = sizes[1];
+    H = 1;
+    W = sizes[2];
+  } else {
+    N = 1;
+    C = sizes[length - 3];
+    H = sizes[length - 2];
+    W = sizes[length - 1];
+    for (size_t i = 0; i < length - 3; i++) {
+      N *= sizes[i];
+    }
   }
   switch (element_size) {
     case 1:
@@ -252,7 +285,8 @@ bool multipleChannelsPresent(const ArrayRef<exec_aten::SizesType>& sizes) {
   if (length < 3) {
     return true;
   }
-  exec_aten::SizesType C = sizes[length - 3];
+  // For 3D [N, C, W], channel dim is sizes[1]. For 4D+, it's sizes[length-3].
+  exec_aten::SizesType C = (length == 3) ? sizes[1] : sizes[length - 3];
   return C != 1;
 }
 
@@ -431,7 +465,7 @@ class NeutronBackend final : public PyTorchBackendInterface {
           return Error::InvalidProgram;
         }
 
-        if (is_channels_last_dim_order(dim_order, arg.dim())) {
+        if (is_channels_last_dim_order_extended(dim_order, arg.dim())) {
           // The tensor is already permuted.
           ET_LOG(Debug, "Using channels last dim order for input %d.\n", i);
           cfg->dcfg.inputs[i] = arg.const_data_ptr();
@@ -481,7 +515,7 @@ class NeutronBackend final : public PyTorchBackendInterface {
           multipleChannelsPresent(arg.sizes())) {
         // The output will have to be transposed.
 
-        if (is_channels_last_dim_order(dim_order, arg.dim())) {
+        if (is_channels_last_dim_order_extended(dim_order, arg.dim())) {
           // The tensor will already be correctly permuted. No transposition
           //  needed.
           cfg->dcfg.outputs[i] = arg.mutable_data_ptr();
@@ -539,7 +573,7 @@ class NeutronBackend final : public PyTorchBackendInterface {
         }
 
         auto dim_order = arg.dim_order().data();
-        if (is_channels_last_dim_order(dim_order, arg.dim())) {
+        if (is_channels_last_dim_order_extended(dim_order, arg.dim())) {
           // The rest of the model expects the `channels_last` dim order, which
           //  the data already matches.
           ET_LOG(Debug, "Using channels last dim order for output %d.\n", i);
@@ -587,4 +621,4 @@ static auto registered = register_backend(backend_id);
 } // namespace
 } // namespace neutron
 } // namespace executor
-} // namespace torch
\ No newline at end of file
+} // namespace torch
diff --git a/backends/nxp/tests/executorch_pipeline.py b/backends/nxp/tests/executorch_pipeline.py
index 8f588be621d..bd6edc8afa3 100644
--- a/backends/nxp/tests/executorch_pipeline.py
+++ b/backends/nxp/tests/executorch_pipeline.py
@@ -164,9 +164,9 @@ def _get_example_input(
             case torch.contiguous_format:
                 sample = torch.ones(spec.shape, dtype=spec.dtype)
             case torch.channels_last:
-                sample = torch.ones(spec.shape, dtype=spec.dtype).to(
-                    memory_format=torch.channels_last
-                )
+                sample = torch.ones(spec.shape, dtype=spec.dtype)
+                if len(spec.shape) >= 4:
+                    sample = sample.to(memory_format=torch.channels_last)
             case _:
                 raise ValueError(f"Unsupported dim_order: {spec.dim_order}")
         # noinspection PyUnboundLocalVariable
diff --git a/backends/nxp/tests/generic_tests/test_3d_channels_last.py b/backends/nxp/tests/generic_tests/test_3d_channels_last.py
new file mode 100644
index 00000000000..837bae3bee0
--- /dev/null
+++ b/backends/nxp/tests/generic_tests/test_3d_channels_last.py
@@ -0,0 +1,46 @@
+# Copyright 2026 NXP
+#
+# This source code is licensed under the BSD-style license found in the
+# LICENSE file in the root directory of this source tree.
+
+import torch
+
+from executorch.backends.nxp.tests.dataset_creator import RandomDatasetCreator
+from executorch.backends.nxp.tests.executorch_pipeline import ModelInputSpec
+from executorch.backends.nxp.tests.graph_verifier import (
+    BaseGraphVerifier,
+    NonDelegatedNode,
+)
+from executorch.backends.nxp.tests.model_output_comparator import (
+    NumericalStatsOutputComparator,
+)
+from executorch.backends.nxp.tests.nsys_testing import lower_run_compare, ReferenceModel
+
+
+class Conv1DModel(torch.nn.Module):
+
+    def __init__(self):
+        super().__init__()
+        self.conv = torch.nn.Conv1d(2, 2, 3)
+
+    def forward(self, x):
+        return self.conv(x)
+
+
+def test_3d_channels_last_dim_order__conv(mocker):
+    model = Conv1DModel().eval()
+
+    input_spec = [ModelInputSpec((1, 2, 5), dim_order=torch.channels_last)]
+
+    comparator = NumericalStatsOutputComparator(max_mse_error=8e-3)
+    lower_run_compare(
+        model,
+        input_spec,
+        dataset_creator=RandomDatasetCreator(),
+        output_comparator=comparator,
+        dlg_model_verifier=BaseGraphVerifier(
+            1, [NonDelegatedNode("aten_view_copy_default", 2)]
+        ),
+        mocker=mocker,
+        reference_model=ReferenceModel.QUANTIZED_EDGE_PYTHON,
+    )
diff --git a/backends/nxp/tests/nsys_testing.py b/backends/nxp/tests/nsys_testing.py
index 9d25c309d25..283ba317a51 100644
--- a/backends/nxp/tests/nsys_testing.py
+++ b/backends/nxp/tests/nsys_testing.py
@@ -334,7 +334,9 @@ def _run_python_program(
                         list(spec.shape)
                     )
                     sample = np.moveaxis(sample.reshape(channels_last_shape), -1, 1)
-                    sample = torch.tensor(sample).to(memory_format=torch.channels_last)
+                    sample = torch.tensor(sample)
+                    if len(spec.shape) >= 4:
+                        sample = sample.to(memory_format=torch.channels_last)
 
                 case _:
                     raise ValueError(f"Unsupported dim_order: {spec.dim_order}")