diff --git a/mindspore/ops/_op_impl/aicpu/__init__.py b/mindspore/ops/_op_impl/aicpu/__init__.py
index 475200e7ef87d3c762f9d05b7695d97fe8bdb804..24f4ad750cda85df9156b9bf4ec06f3787fff14b 100644
--- a/mindspore/ops/_op_impl/aicpu/__init__.py
+++ b/mindspore/ops/_op_impl/aicpu/__init__.py
@@ -25,3 +25,4 @@ from .squeeze import _squeeze_aicpu
 from .expand_dims import _expand_dims_aicpu
 from .random_choice_with_mask import _random_choice_with_mask_aicpu
 from .pack import _pack_aicpu
+from .normal import _normal_aicpu
diff --git a/mindspore/ops/_op_impl/aicpu/normal.py b/mindspore/ops/_op_impl/aicpu/normal.py
new file mode 100644
index 0000000000000000000000000000000000000000..fdb96e362fd0e4048f52d1bc2c0697aeff004c56
--- /dev/null
+++ b/mindspore/ops/_op_impl/aicpu/normal.py
@@ -0,0 +1,33 @@
+# Copyright 2020 Huawei Technologies Co., Ltd
+#
+# 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.
+# ============================================================================
+
+"""Normal op"""
+from mindspore.ops.op_info_register import op_info_register, AiCPURegOp, DataType
+
+normal_op_info = AiCPURegOp("Normal") \
+    .fusion_type("OPAQUE") \
+    .input(0, "shape", "required") \
+    .input(1, "mean", "required") \
+    .input(2, "stddev", "required") \
+    .output(0, "y", "required") \
+    .attr("seed", "int") \
+    .dtype_format(DataType.I32_Default, DataType.F32_Default, DataType.F32_Default, DataType.F32_Default) \
+    .dtype_format(DataType.I32_NCHW, DataType.F32_NCHW, DataType.F32_NCHW, DataType.F32_NCHW) \
+    .get_op_info()
+
+@op_info_register(normal_op_info)
+def _normal_aicpu():
+    """Normal AiCPU register"""
+    return
diff --git a/mindspore/ops/operations/__init__.py b/mindspore/ops/operations/__init__.py
index f0ac503604846da84ab2b3b9a0382aac8e4a9412..57ffd969c14e8fa1fdf545cc5a0752df4f0cdf0f 100644
--- a/mindspore/ops/operations/__init__.py
+++ b/mindspore/ops/operations/__init__.py
@@ -53,7 +53,7 @@ from .math_ops import (Abs, ACos, Asin, Asinh, AddN, AssignAdd, AssignSub, Atan2
                        Sin, Sqrt, Rsqrt, BesselI0e, BesselI1e,
                        Square, Sub, TensorAdd, Sign, Round, SquareSumAll, Atan, Atanh, Cosh, Sinh)
 
-from .random_ops import (RandomChoiceWithMask)
+from .random_ops import (RandomChoiceWithMask, Normal)
 from .nn_ops import (LSTM, SGD, Adam, SparseApplyAdam, SparseApplyLazyAdam, ApplyMomentum, BatchNorm,
                      BiasAdd, Conv2D,
                      DepthwiseConv2dNative,
@@ -163,6 +163,7 @@ __all__ = [
     'HSigmoid',
     'Tanh',
     'RandomChoiceWithMask',
+    'Normal',
     'ResizeBilinear',
     'ScalarSummary',
     'ImageSummary',
diff --git a/mindspore/ops/operations/random_ops.py b/mindspore/ops/operations/random_ops.py
index 2692b43b46c7aa7bbe268a995964cb0097ccc2d1..7a457d099818b726b49f691b7cd4099abd7ca055 100644
--- a/mindspore/ops/operations/random_ops.py
+++ b/mindspore/ops/operations/random_ops.py
@@ -64,3 +64,47 @@ class RandomChoiceWithMask(PrimitiveWithInfer):
     def infer_dtype(self, x_dtype):
         validator.check_tensor_type_same({'x': x_dtype}, [mstype.bool_], self.name)
         return (mstype.int32, mstype.bool_)
+
+
+class Normal(PrimitiveWithInfer):
+    """
+    Generates random samples from a normal(Gaussian) distribution.
+
+    Args:
+        seed (int): Random seed. Default: 0.
+
+    Inputs:
+        - **shape** (tuple[int]) - The shape of output tensor. Only constant value is allowed.
+        - **mean** (Tensor) - The mean of the distribution, with float32 data type.
+        - **stddev** (Tensor) - The standard deviation of the distribution, with float32 data type.
+
+    Outputs:
+        Tensor, with the given shape from the specific distribution and float32 data type.
+
+    Examples:
+        >>> normal = P.Normal()
+        >>> mean = Tensor(0., mstype.float32)
+        >>> stddev = Tensor(1., mstype.float32)
+        >>> out = normal((32, 3, 3), mean, stddev)
+    """
+
+    @prim_attr_register
+    def __init__(self, seed=0):
+        """Init Normal"""
+        validator.check_value_type("seed", seed, [int], self.name)
+
+    def __infer__(self, shape, mean, stddev):
+        shape_value = shape["value"]
+        if shape_value is None:
+            raise ValueError(f"For {self.name}, shape must be const.")
+        validator.check_value_type("shape", shape_value, [tuple], self.name)
+        for i, shape_i in enumerate(shape_value):
+            validator.check_integer("shape[%d]" % i, shape_i, 0, Rel.GE, self.name)
+
+        validator.check_tensor_type_same({"mean": mean["dtype"]}, [mstype.float32], self.name)
+        validator.check_tensor_type_same({"stddev": stddev["dtype"]}, [mstype.float32], self.name)
+
+        out = {"shape": shape_value,
+               "dtype": mstype.float32,
+               "value": None}
+        return out
diff --git a/tests/st/ops/ascend/test_aicpu_ops/test_normal.py b/tests/st/ops/ascend/test_aicpu_ops/test_normal.py
new file mode 100644
index 0000000000000000000000000000000000000000..66254caf21f162e55af424bfd1d3739fffb2c8b4
--- /dev/null
+++ b/tests/st/ops/ascend/test_aicpu_ops/test_normal.py
@@ -0,0 +1,43 @@
+# Copyright 2020 Huawei Technologies Co., Ltd
+#
+# 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.
+# ============================================================================
+import mindspore.context as context
+import mindspore.nn as nn
+from mindspore.ops import operations as P
+from mindspore.common import Tensor
+from mindspore.common import dtype as mstype
+
+
+context.set_context(mode=context.PYNATIVE_MODE, device_target="Ascend")
+
+
+class Net(nn.Cell):
+    def __init__(self, shape=None, mean=0.0, stddev=1.0, seed=0):
+        super(Net, self).__init__()
+        self._mean = Tensor(mean, mstype.float32)
+        self._stddev = Tensor(stddev, mstype.float32)
+        self._normal = P.Normal(seed=seed)
+        self._shape = shape
+
+    def construct(self):
+        return self._normal(self._shape, self._mean, self._stddev)
+
+
+def test_net_3x2x4():
+    mean = 0.0
+    stddev = 1.0
+    seed = 0
+    net = Net((3, 2, 4), mean, stddev, seed)
+    out = net()
+    assert out.shape == (3, 2, 4)
diff --git a/tests/ut/python/ops/test_ops.py b/tests/ut/python/ops/test_ops.py
index ad18faa61ee8571dac2295384bd9a93671d85706..5927d97c50b54a8b765b6b9780f8f5fdecb6632b 100755
--- a/tests/ut/python/ops/test_ops.py
+++ b/tests/ut/python/ops/test_ops.py
@@ -399,6 +399,19 @@ class InplaceSubNet(nn.Cell):
         return out
 
 
+class NormalNet(nn.Cell):
+    def __init__(self, shape=None, mean=0.0, stddev=1.0, seed=0):
+        super(NormalNet, self).__init__()
+        self.normal = P.Normal(seed=seed)
+        self.shape = shape
+        self.mean = Tensor(mean, mstype.float32)
+        self.stddev = Tensor(stddev, mstype.float32)
+
+    def construct(self):
+        out = self.normal(self.shape, self.mean, self.stddev)
+        return out
+
+
 test_case_math_ops = [
     ('BitwiseAnd', {
         'block': P.BitwiseAnd(),
@@ -895,6 +908,10 @@ test_case_math_ops = [
         'desc_inputs': [Tensor([-1.0, 0.0, 1.5, 2.0, 5.0, 15], mstype.float16), Tensor([0.0, 5.0], mstype.float16)],
         'desc_bprop': [],
         'skip': ['backward']}),
+    ('Normal', {
+        'block': NormalNet((3, 2, 4), 0.0, 1.0, 0),
+        'desc_inputs': [],
+        'skip': ['backward']}),
 ]
 
 test_case_nn_ops = [