diff --git a/src/backends/reference/workloads/RefSoftmaxWorkload.cpp b/src/backends/reference/workloads/RefSoftmaxWorkload.cpp
new file mode 100644
index 0000000..b176667
--- /dev/null
+++ b/src/backends/reference/workloads/RefSoftmaxWorkload.cpp
@@ -0,0 +1,39 @@
+﻿//
+// Copyright © 2017 Arm Ltd. All rights reserved.
+// SPDX-License-Identifier: MIT
+//
+
+#include "RefSoftmaxWorkload.hpp"
+
+#include "Decoders.hpp"
+#include "Encoders.hpp"
+#include "RefWorkloadUtils.hpp"
+#include "Softmax.hpp"
+
+#include "Profiling.hpp"
+
+#include <vector>
+
+namespace armnn
+{
+
+void RefSoftmaxWorkload::Execute() const
+{
+    ARMNN_SCOPED_PROFILING_EVENT(Compute::CpuRef, "RefSoftmaxWorkload_Execute");
+
+    const TensorInfo &inputTensorInfo = GetTensorInfo(m_Data.m_Inputs[0]);
+
+    std::unique_ptr<Decoder<float>> decoderPtr = MakeDecoder<float>(inputTensorInfo, m_Data.m_Inputs[0]->Map());
+    Decoder<float> &decoder = *decoderPtr;
+
+    const TensorInfo &outputTensorInfo = GetTensorInfo(m_Data.m_Outputs[0]);
+
+    std::unique_ptr<Encoder<float>> encoderPtr = MakeEncoder<float>(outputTensorInfo, m_Data.m_Outputs[0]->Map());
+    Encoder<float> &encoder = *encoderPtr;
+
+    Softmax(decoder,
+            encoder,
+            inputTensorInfo,
+            m_Data.m_Parameters.m_Beta);
+}
+} //namespace armnn