diff --git a/src/runtime/CL/functions/CLGEMM.cpp b/src/runtime/CL/functions/CLGEMM.cpp
index 0c82e6d..6adbdc0 100644
--- a/src/runtime/CL/functions/CLGEMM.cpp
+++ b/src/runtime/CL/functions/CLGEMM.cpp
@@ -133,10 +133,7 @@
     if(_is_interleaved_transposed)
     {
         reinterpret_input_as_3d = false;
-    }
 
-    if(_is_interleaved_transposed)
-    {
         matrix_a = &_tmp_a;
         matrix_b = &_tmp_b;
 
@@ -200,8 +197,7 @@
     // If we pass the matrix A and matrix B reshaped to CLGEMMMatrixMultiplyKernel, we need to pass m, n, k, mult_transpose1xW_width and mult_interleave4x4_height to CLGEMMReshapeInfo
     // in order to know how the matrices have been reshaped
     bool      reinterpret_input_as_3d   = gemm_info.reinterpret_input_as_3d();
-    bool      reinterpret_output_as_3d  = (gemm_info.depth_output_gemm3d() != 1);
-    const int m                         = a->dimension(1);
+    const int m                         = reinterpret_input_as_3d ? (a->dimension(1) * a->dimension(2)) : a->dimension(1);
     const int n                         = b->dimension(0);
     const int k                         = a->dimension(0);
     int       mult_transpose1xW_width   = 1;
@@ -217,21 +213,13 @@
     // Check if we need to reshape the matrix A and matrix B
     const bool run_interleave_transpose = is_interleaved_transposed(m, n, k, a->data_type(), reshape_b_only_on_first_run, gpu_target);
 
-    // In case both input and output have to be reinterpreted as 3D tensors,
-    // force reinterpret_input_as_3d and reinterpret_output_as_3d to be false.
-    if(reinterpret_input_as_3d == reinterpret_output_as_3d)
-    {
-        reinterpret_input_as_3d  = false;
-        reinterpret_output_as_3d = false;
-    }
-
     // if _is_interleaved_transposed is set, force reinterpret_input_as_3d to be false as the output of CLGEMMInterleaveKernel will be 2D
     if(run_interleave_transpose)
     {
         reinterpret_input_as_3d = false;
     }
 
-    const GEMMReshapeInfo reshape_info = GEMMReshapeInfo(m, n, k, mult_transpose1xW_width, mult_interleave4x4_height, reinterpret_output_as_3d ? depth_output_gemm3d : 1, reinterpret_input_as_3d);
+    const GEMMReshapeInfo reshape_info = GEMMReshapeInfo(m, n, k, mult_transpose1xW_width, mult_interleave4x4_height, depth_output_gemm3d, reinterpret_input_as_3d);
 
     if(run_interleave_transpose)
     {
@@ -239,8 +227,8 @@
         matrix_b_info = &tmp_b_info;
 
         // Validate interleave kernel
-        auto_init_if_empty(tmp_a_info, a->clone()->set_tensor_shape(compute_interleaved_shape(*a, mult_interleave4x4_height, reinterpret_input_as_3d)));
-        ARM_COMPUTE_RETURN_ON_ERROR(CLGEMMInterleave4x4Kernel::validate(a, &tmp_a_info, mult_interleave4x4_height, reinterpret_input_as_3d));
+        auto_init_if_empty(tmp_a_info, a->clone()->set_tensor_shape(compute_interleaved_shape(*a, mult_interleave4x4_height, gemm_info.reinterpret_input_as_3d())));
+        ARM_COMPUTE_RETURN_ON_ERROR(CLGEMMInterleave4x4Kernel::validate(a, &tmp_a_info, mult_interleave4x4_height, gemm_info.reinterpret_input_as_3d()));
 
         // Validate transpose kernel
         auto_init_if_empty(tmp_b_info, b->clone()->set_tensor_shape(compute_transpose1xW_with_element_size_shape(*b, mult_transpose1xW_width)));
diff --git a/src/runtime/CL/functions/CLGEMMLowpMatrixMultiplyCore.cpp b/src/runtime/CL/functions/CLGEMMLowpMatrixMultiplyCore.cpp
index 7aeade1..509b668 100644
--- a/src/runtime/CL/functions/CLGEMMLowpMatrixMultiplyCore.cpp
+++ b/src/runtime/CL/functions/CLGEMMLowpMatrixMultiplyCore.cpp
@@ -108,22 +108,13 @@
     // If we pass the matrix A and matrix B reshaped to CLGEMMMatrixMultiplyKernel, we need to pass m, n, k, mult_transpose1xW_width and mult_interleave4x4_height to CLGEMMReshapeInfo
     // in order to know how the matrices have been reshaped
     bool          reinterpret_input_as_3d   = gemm_info.reinterpret_input_as_3d();
-    bool          reinterpret_output_as_3d  = (gemm_info.depth_output_gemm3d() != 1);
-    int           m                         = reinterpret_input_as_3d ? (a->info()->dimension(1) * a->info()->dimension(2)) : a->info()->dimension(1);
+    const int     m                         = reinterpret_input_as_3d ? (a->info()->dimension(1) * a->info()->dimension(2)) : a->info()->dimension(1);
     const int     n                         = b->info()->dimension(0);
     const int     k                         = a->info()->dimension(0);
     const int     depth_output_gemm3d       = gemm_info.depth_output_gemm3d();
     constexpr int mult_transpose1xW_width   = 1;
     constexpr int mult_interleave4x4_height = 1;
 
-    // In case both input and output have to be reinterpreted as 3D tensors,
-    // force reinterpret_input_as_3d and reinterpret_output_as_3d to be false.
-    if(reinterpret_input_as_3d == reinterpret_output_as_3d)
-    {
-        reinterpret_input_as_3d  = false;
-        reinterpret_output_as_3d = false;
-    }
-
     // Check if we need to reshape the matrix A and matrix B
     _is_interleaved_transposed = is_interleaved_transposed(m, n, k, _reshape_b_only_on_first_run, gpu_target);
 
@@ -131,7 +122,6 @@
     {
         // if _is_interleaved_transposed is set, force reinterpret_input_as_3d to be false as the output of CLGEMMInterleaveKernel will be 2D
         reinterpret_input_as_3d = false;
-        m                       = a->info()->dimension(1);
 
         matrix_a = &_tmp_a;
         matrix_b = &_tmp_b;
@@ -143,7 +133,7 @@
         }
 
         // Configure interleave kernel
-        _mtx_a_reshape_kernel.configure(a, &_tmp_a, mult_interleave4x4_height, reinterpret_input_as_3d);
+        _mtx_a_reshape_kernel.configure(a, &_tmp_a, mult_interleave4x4_height, gemm_info.reinterpret_input_as_3d());
 
         // Configure transpose kernel
         _mtx_b_reshape_kernel.configure(b, &_tmp_b, mult_transpose1xW_width);
@@ -151,7 +141,7 @@
     // Configure matrix multiply kernel
     _mm_kernel.configure(matrix_a, matrix_b, output, _is_interleaved_transposed, GEMMReshapeInfo(m, n, k,
                                                                                                  mult_transpose1xW_width, mult_interleave4x4_height,
-                                                                                                 reinterpret_output_as_3d ? depth_output_gemm3d : 1, reinterpret_input_as_3d));
+                                                                                                 depth_output_gemm3d, reinterpret_input_as_3d));
 
     // Initialize matrix B reduction kernel only if _a_offset is not equal to 0
     if(_a_offset != 0)
@@ -213,23 +203,20 @@
     int32_t a_offset = a->quantization_info().offset;
     int32_t b_offset = b->quantization_info().offset;
 
+    const ITensorInfo *matrix_a_info = a;
+    const ITensorInfo *matrix_b_info = b;
+
+    TensorInfo tmp_a_info{};
+    TensorInfo tmp_b_info{};
+
     bool          reinterpret_input_as_3d   = gemm_info.reinterpret_input_as_3d();
-    bool          reinterpret_output_as_3d  = (gemm_info.depth_output_gemm3d() != 1);
-    int           m                         = a->dimension(1);
+    const int     m                         = reinterpret_input_as_3d ? (a->dimension(1) * a->dimension(2)) : a->dimension(1);
     const int     n                         = b->dimension(0);
     const int     k                         = a->dimension(0);
     constexpr int mult_transpose1xW_width   = 1;
     constexpr int mult_interleave4x4_height = 1;
     const int     depth_output_gemm3d       = gemm_info.depth_output_gemm3d();
 
-    // In case both input and output have to be reinterpreted as 3D tensors,
-    // force reinterpret_input_as_3d and reinterpret_output_as_3d to be false.
-    if(reinterpret_input_as_3d == reinterpret_output_as_3d)
-    {
-        reinterpret_input_as_3d  = false;
-        reinterpret_output_as_3d = false;
-    }
-
     bool reshape_matrices = is_interleaved_transposed(m, n, k, gemm_info.reshape_b_only_on_first_run(), CLScheduler::get().target());
 
     // if reshape_matrices is set, force reinterpret_input_as_3d to be false as the output of CLGEMMInterleaveKernel will be 2D
@@ -238,21 +225,23 @@
         reinterpret_input_as_3d = false;
     }
 
-    const GEMMReshapeInfo reshape_info = GEMMReshapeInfo(m, n, k, mult_transpose1xW_width, mult_interleave4x4_height, reinterpret_output_as_3d ? depth_output_gemm3d : 1, reinterpret_input_as_3d);
+    const GEMMReshapeInfo reshape_info = GEMMReshapeInfo(m, n, k, mult_transpose1xW_width, mult_interleave4x4_height, depth_output_gemm3d, reinterpret_input_as_3d);
 
     if(reshape_matrices)
     {
-        TensorInfo info_a(compute_interleaved_shape(*a, mult_interleave4x4_height, reinterpret_input_as_3d), 1, a->data_type());
-        TensorInfo info_b(compute_transpose1xW_with_element_size_shape(*b, mult_transpose1xW_width), 1, b->data_type());
+        matrix_a_info = &tmp_a_info;
+        matrix_b_info = &tmp_b_info;
 
-        ARM_COMPUTE_RETURN_ON_ERROR(CLGEMMInterleave4x4Kernel::validate(a, &info_a, mult_interleave4x4_height, reinterpret_input_as_3d));
-        ARM_COMPUTE_RETURN_ON_ERROR(CLGEMMTranspose1xWKernel::validate(b, &info_b, mult_transpose1xW_width));
-        ARM_COMPUTE_RETURN_ON_ERROR(CLGEMMLowpMatrixMultiplyKernel::validate(&info_a, &info_b, output, reshape_matrices, reshape_info));
+        // Validate interleave kernel
+        auto_init_if_empty(tmp_a_info, a->clone()->set_tensor_shape(compute_interleaved_shape(*a, mult_interleave4x4_height, gemm_info.reinterpret_input_as_3d())));
+        ARM_COMPUTE_RETURN_ON_ERROR(CLGEMMInterleave4x4Kernel::validate(a, &tmp_a_info, mult_interleave4x4_height, gemm_info.reinterpret_input_as_3d()));
+
+        // Validate transpose kernel
+        auto_init_if_empty(tmp_b_info, b->clone()->set_tensor_shape(compute_transpose1xW_with_element_size_shape(*b, mult_transpose1xW_width)));
     }
-    else
-    {
-        ARM_COMPUTE_RETURN_ON_ERROR(CLGEMMLowpMatrixMultiplyKernel::validate(a, b, output, reshape_matrices, reshape_info));
-    }
+
+    // Validate matrix multiply
+    ARM_COMPUTE_RETURN_ON_ERROR(CLGEMMLowpMatrixMultiplyKernel::validate(matrix_a_info, matrix_b_info, output, reshape_matrices, reshape_info));
 
     TensorInfo info_vector_sum_col, info_vector_sum_row;