diff --git a/paddle/fluid/operators/cast_op_npu.cc b/paddle/fluid/operators/cast_op_npu.cc
new file mode 100755
index 0000000000000000000000000000000000000000..5cfb152684e9a1d2a62207429f225e0a1217be48
--- /dev/null
+++ b/paddle/fluid/operators/cast_op_npu.cc
@@ -0,0 +1,91 @@
+/* Copyright (c) 2021 PaddlePaddle Authors. All Rights Reserved.
+
+Licensed under the Apache License, Version 2.0 (the "License");
+you may not use this file except in compliance with the License.
+You may obtain a copy of the License at
+
+    http://www.apache.org/licenses/LICENSE-2.0
+
+Unless required by applicable law or agreed to in writing, software
+distributed under the License is distributed on an "AS IS" BASIS,
+WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+See the License for the specific language governing permissions and
+limitations under the License. */
+
+#ifdef PADDLE_WITH_ASCEND_CL
+#include <memory>
+#include <string>
+
+#include "paddle/fluid/operators/cast_op.h"
+#include "paddle/fluid/operators/npu_op_runner.h"
+
+namespace paddle {
+namespace operators {
+
+static std::map<framework::proto::VarType::Type, aclDataType>
+    DTYPE_2_ACL_DTYPE = {
+        {framework::proto::VarType::BOOL, ACL_BOOL},
+        {framework::proto::VarType::INT16, ACL_INT16},
+        {framework::proto::VarType::INT32, ACL_INT32},
+        {framework::proto::VarType::INT64, ACL_INT64},
+        {framework::proto::VarType::FP16, ACL_FLOAT16},
+        {framework::proto::VarType::FP32, ACL_FLOAT},
+        {framework::proto::VarType::FP64, ACL_DOUBLE},
+};
+
+using Tensor = framework::Tensor;
+
+template <typename DeviceContext, typename T>
+class CastNPUKernel : public framework::OpKernel<T> {
+ public:
+  void Compute(const framework::ExecutionContext& ctx) const override {
+    auto* x = ctx.Input<Tensor>("X");
+    int dtype = ctx.Attr<int>("out_dtype");
+
+    auto* out = ctx.Output<Tensor>("Out");
+
+    auto place = ctx.GetPlace();
+   
+    auto iter = DTYPE_2_ACL_DTYPE.find(static_cast<framework::proto::VarType::Type>(dtype));
+    int aclDtype = iter->second;
+
+    if (dtype == framework::proto::VarType::FP32) {
+        out->mutable_data<float>(place);
+    } else if (dtype == framework::proto::VarType::FP16) {
+        out->mutable_data<paddle::platform::float16>(place);
+    } else if (dtype == framework::proto::VarType::INT16) {
+        out->mutable_data<int16_t>(place);
+    } else if (dtype == framework::proto::VarType::INT32) {
+        out->mutable_data<int32_t>(place);
+    } else if (dtype == framework::proto::VarType::INT64) {
+        out->mutable_data<int64_t>(place);
+    } else if (dtype == framework::proto::VarType::FP64) {
+        out->mutable_data<double>(place);
+    } else if (dtype == framework::proto::VarType::BOOL) {
+        out->mutable_data<bool>(place);
+    }
+
+    auto stream =
+        ctx.template device_context<paddle::platform::NPUDeviceContext>()
+            .stream();
+
+    auto runner = NpuOpRunner("Cast", {*x}, {*out}, {{"dst_type", static_cast<int32_t>(aclDtype)}});
+    runner.Run(stream);
+  }
+};
+}  // namespace operators
+}  // namespace paddleaclDtype
+
+namespace ops = paddle::operators;
+
+REGISTER_OP_NPU_KERNEL(
+    cast,
+    ops::CastNPUKernel<paddle::platform::NPUDeviceContext, int16_t>,
+    ops::CastNPUKernel<paddle::platform::NPUDeviceContext, int32_t>,
+    ops::CastNPUKernel<paddle::platform::NPUDeviceContext, int64_t>,
+    ops::CastNPUKernel<paddle::platform::NPUDeviceContext, bool>,
+    ops::CastNPUKernel<paddle::platform::NPUDeviceContext, double>,
+    ops::CastNPUKernel<paddle::platform::NPUDeviceContext, float>,
+    ops::CastNPUKernel<paddle::platform::NPUDeviceContext,
+    paddle::platform::float16>);
+#endif
diff --git a/python/paddle/fluid/tests/unittests/npu/test_cast_op_npu.py b/python/paddle/fluid/tests/unittests/npu/test_cast_op_npu.py
new file mode 100755
index 0000000000000000000000000000000000000000..97a10f6657617dcfd6c57d5bc9b9ca2b142f9e7e
--- /dev/null
+++ b/python/paddle/fluid/tests/unittests/npu/test_cast_op_npu.py
@@ -0,0 +1,75 @@
+#  Copyright (c) 2021 PaddlePaddle Authors. All Rights Reserved.
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#    http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+
+from __future__ import print_function
+
+import numpy as np
+import unittest
+import sys
+sys.path.append("..")
+from op_test import OpTest
+import paddle
+import paddle.fluid as fluid
+import paddle.fluid.core as core
+
+paddle.enable_static()
+SEED = 2021
+
+
+@unittest.skipIf(not paddle.is_compiled_with_npu(),
+                 "core is not compiled with NPU")
+class TestCast1(OpTest):
+    def setUp(self):
+        self.set_npu()
+        self.op_type = "cast"
+        self.place = paddle.NPUPlace(0)
+
+        ipt = np.random.random(size=[10, 10]) + 1
+        self.inputs = {'X': ipt.astype('float32')}
+        self.outputs = {'Out': ipt.astype('float16')}
+
+        self.attrs = {
+            'in_dtype': int(core.VarDesc.VarType.FP32),
+            'out_dtype': int(core.VarDesc.VarType.FP16)
+        }
+
+    def set_npu(self):
+        self.__class__.use_npu = True
+
+    def test_check_output(self):
+        self.check_output_with_place(self.place, check_dygraph=False)
+
+class TestCast2(OpTest):
+    def setUp(self):
+        self.set_npu()
+        self.op_type = "cast"
+        self.place = paddle.NPUPlace(0)
+
+        ipt = np.random.random(size=[10, 10]) + 1
+        self.inputs = {'X': ipt.astype('float16')}
+        self.outputs = {'Out': ipt.astype('float32')}
+
+        self.attrs = {
+            'in_dtype': int(core.VarDesc.VarType.FP16),
+            'out_dtype': int(core.VarDesc.VarType.FP32)
+        }
+
+    def set_npu(self):
+        self.__class__.use_npu = True
+
+    def test_check_output(self):
+        self.check_output_with_place(self.place, check_dygraph=False, atol=1e-3)
+
+if __name__ == '__main__':
+    unittest.main()